No entries match these filters. Clear filters
Arcade adds scoped and deactivatable project API keys, per-gateway Registry catalog access, named OAuth connection profiles, and Client ID Metadata Document Client ID Metadata Document An OAuth 2.0 mechanism letting a client publish its own metadata (redirect URIs, grant types, contacts) at a well-known URL so authorization servers can fetch and verify it without manual registration. auth for remote M︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄
└──▷ USE IT
Declare additional Debian packages your deployed server needs at build time, so they are available on first pod start without recompiling.
.arcade/deploy.json
copy
{
"apt_packages": ["libpq-dev", "libxml2"]
}
Prevent a CI pipeline or runaway automation from flooding a project's deployment queue by capping how many deployments can be pending at once.
yaml
copy
deployments:
max_pending_per_project: 5
Deployed servers can declare apt_packages in .arcade/deploy.json to install Debian packages at build time.︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄ Not checked yet Cap pending deployments per project on self-hosted installs with deployments.max_pending_per_project in engine.extraConfig. Not checked yet A gateway can expose two Registry tools letting a signed-in agent browse the full project MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. catalog, behind a per-gateway setting off by default. Not checked yet Tool execution results now report a new error kind for deliberate tool refusals (bad input or business rules); the prior fatal-error kind is now reserved for unhandled crashes. Not checked yet (improved) Several toolkit integrations gain upgrades: Google Drive's query parameter is now optional and searches every shared drive by default, flags incomplete results, and reports last editor and sharer; Google Docs' EditDocument can now return only named tabs instead of the whole document; Gmail's GetThread takes a flag to leave out each message's HTML body; Linear adds server-side filtering for ListInitiatives and ListProjectComments now returns discussion comments too; and GitHub's DownloadFile on a directory now explains and lists contents, with GraphQL tools now working on GitHub Enterprise Server. Not checked yet
+5 more
show less
Admins creating a project API key now choose its scope — call tools, manage the project, or both — instead of always granting full access. Not checked yet Admins can now deactivate a project API key immediately without deleting it, and reactivate it without issuing a new secret. Not checked yet Organizations can turn on named connection profiles per OAuth provider so an end-user can switch accounts for the same toolkit. Not checked yet Authorization requests and responses now record an expected issuer, supporting Client ID Metadata Document authorization for remote MCP servers. Not checked yet A gateway can now link a remote MCP App's resources back to the gateway that served them. Not checked yet
Deployed servers can declare apt_packages in .arcade/deploy.json to install Debian packages at build time.︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄ Not checked yet Cap pending deployments per project on self-hosted installs with deployments.max_pending_per_project in engine.extraConfig. Not checked yet A gateway can expose two Registry tools letting a signed-in agent browse the full project MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. catalog, behind a per-gateway setting off by default. Not checked yet Tool execution results now report a new error kind for deliberate tool refusals (bad input or business rules); the prior fatal-error kind is now reserved for unhandled crashes. Not checked yet Admins can now deactivate a project API key immediately without deleting it, and reactivate it without issuing a new secret. Not checked yet
+3 more
show less
Organizations can turn on named connection profiles per OAuth provider so an end-user can switch accounts for the same toolkit. Not checked yet Authorization requests and responses now record an expected issuer, supporting Client ID Metadata Document authorization for remote MCP servers. Not checked yet A gateway can now link a remote MCP App's resources back to the gateway that served them. Not checked yet
2 more in Everything
└──▷ BREAKING ON UPGRADE
! The tool-execution-history API key/gateway columns are now nullable and required — existing integrations must handle null values.! The old account-based gateway binding is dropped; any gateway still using it will be blocked until migrated (see Upgrade notes).! The porter deployment provider is removed (see Upgrade notes).! The runner image's preinstalled unixODBC, libaio, GSSAPI, and Oracle Instant Client compatibility link are removed from the deployed-server runner image (see Upgrade notes).
Arcade adds batch tool-execution history deletion, paged Google Drive search, and new GitHub, Linear, Slack, and Teams tools with per-execution API key and gateway reporting︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄
└──▷ USE IT
Avoid accidental overwrites when creating a Word document by explicitly choosing what happens on a name conflict.
yaml
copy
conflict_behavior: rename
Let OpenFGA use a pre-created, least-privilege database instead of trying to run createdb itself.
yaml
copy
rbac:
openfga:
database:
create: false
Improves the Linear toolkit: filters ListIssues results by creator (a specific person or the caller); retrieves completion time, cancellation time, and creator from ListIssues, and completion and cancellation time from GetIssue; lets you clear an issue's parent or project, or set a project's priority; retrieves a sub-issue's parent from ListIssues; and returns each workflow state's team key alongside its ID and name from ListWorkflowStates.︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄ Not checked yet Page Google Drive SearchFiles results past the first page using a continuation token. Not checked yet Control file-conflict behavior in Excel upload and create calls with conflict_behavior (fail, rename, or replace) in place of the previous plain overwrite flag. Not checked yet Record which API key and gateway each tool execution ran through, reported alongside the execution's other details. Not checked yet Delete an organization's stored tool execution history on demand with a new engine command that defaults to a dry run, requires a reason, and caps rows removed per run. Not checked yet
+5 more
show less
Set rbac.openfga.database.create: false so the engine skips running createdb against the postgres maintenance database, letting operators point OpenFGA at a database a DBA already created with a least-privilege role. Not checked yet Adds GitHub tools to search issues and pull requests across repositories using GitHub's search syntax (QueryIssues and QueryPullRequests) and to list every issue and pull request conversation comment in a repository in one paged, filterable and sortable call (ListRepositoryConversationComments). Not checked yet Delete up to 10 Slack files by ID or link in one call with the new DeleteFiles tool, which reports a success, rejection, or unknown outcome per file. Not checked yet List every comment created or edited across a Linear team or project since a given time in one call — including quoted inline comments — with the new ListRecentComments tool. Not checked yet List a Teams meeting's recordings and transcripts in one call with the new ListMeetingArtifacts tool, which reports when an artifact is not available. Not checked yet
Improves the Linear toolkit: filters ListIssues results by creator (a specific person or the caller); retrieves completion time, cancellation time, and creator from ListIssues, and completion and cancellation time from GetIssue; lets you clear an issue's parent or project, or set a project's priority; retrieves a sub-issue's parent from ListIssues; and returns each workflow state's team key alongside its ID and name from ListWorkflowStates.︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄ Not checked yet Record which API key and gateway each tool execution ran through, reported alongside the execution's other details. Not checked yet Delete an organization's stored tool execution history on demand with a new engine command that defaults to a dry run, requires a reason, and caps rows removed per run. Not checked yet Adds GitHub tools to search issues and pull requests across repositories using GitHub's search syntax (QueryIssues and QueryPullRequests) and to list every issue and pull request conversation comment in a repository in one paged, filterable and sortable call (ListRepositoryConversationComments). Not checked yet Delete up to 10 Slack files by ID or link in one call with the new DeleteFiles tool, which reports a success, rejection, or unknown outcome per file. Not checked yet
+2 more
show less
List every comment created or edited across a Linear team or project since a given time in one call — including quoted inline comments — with the new ListRecentComments tool. Not checked yet List a Teams meeting's recordings and transcripts in one call with the new ListMeetingArtifacts tool, which reports when an artifact is not available. Not checked yet
3 more in Everything
└──▷ BREAKING ON UPGRADE
!
Excel upload and create calls: the `overwrite` flag is replaced by `conflict_behavior` (`fail`, `rename`, `replace`) — existing calls using the old flag require migration.
Our best guess
!
Microsoft Word, Excel, and PowerPoint toolkits move to broader file scopes — operator action required on upgrade.
Our best guess
!
Word's create call now refuses a same-named file by default (instead of silently replacing it); the new `conflict_behavior` argument controls the behavior.
Our best guess
!
Figma toolkit: `ListProjects` and `ListFiles` tools are removed — operator action required on upgrade.
Our best guess
Arcade adds Enterprise MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. Hosting so self-hosted customers can deploy custom MCP servers with the arcade deploy command︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄
└──▷ TRY IT
Ship a custom MCP server to your self-hosted Arcade environment from a CI pipeline without filing tickets for a host or cluster access.
Deploy custom MCP servers to self-hosted Arcade environments with the arcade deploy command from a local checkout or CI pipeline, without needing cluster access.︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄ Not checked yet Every deployed MCP server automatically runs behind Arcade's authorization, policy enforcement, and audit logging from its first call. Not checked yet Operators can set deployment limits per organization and per project; when a team hits its limit, the error message states the limit plainly. Not checked yet Monitor deployed servers through the dashboard's Servers page, which shows versioned releases, status, logs, and restart and removal controls. Not checked yet Arcade adds Enterprise MCP Hosting support for Python servers built with the Arcade MCP framework, with FastMCP support planned next. Not checked yet
5 details in Everything
Arcade adds MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. Apps, rendering interactive editors for Gmail, Slack, and Outlook Mail inside the chat, with file upload/download support and policy-enforced UI controls.
Arcade now supports MCP Apps, an extension to the Model Context Protocol Model Context Protocol Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. that renders interactive interfaces (forms, editors, dashboards) inside the chat client instead of requiring tab-switching. Not checked yet New integration File uploads and downloads now work inside MCP Apps by running the app code locally, opening the native file picker and writing large files straight to disk — already available in the Slack, Gmail, and Outlook Mail apps. Not checked yet Authorization and governance rules are enforced at the point when the MCP App interface is built and again at the point of action, so UI controls (such as a Send button) are only shown when the user's permissions allow the underlying tool call. Not checked yet
3 details in Everything
Arcade now lets you register external MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. servers and govern their tools alongside native Arcade tools.︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄
Supports registering an existing or vendor-supplied MCP server with Arcade to govern its tools alongside Arcade's own.︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄ Not checked yet
1 detail in Everything
Arcade Plugin lets agents call 1K+ tools via a single gateway in Claude, Cursor, VS Code, Copilot CLI, and Codex — no MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. config needed.︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄
└──▷ HOW TO FIND IT
Install the Arcade Plugin in a supported agent client to start making tool calls without configuring an MCP gateway.
📍 In your agent client (Claude Code, Claude Desktop, Cursor, VS Code, GitHub Copilot CLI, or Codex): 1. Install the Arcade Plugin from the client's plugin/extension marketplace. 2. Sign in with your Arcade account. 3. Ask your agent something like 'summarize my unread email' or 'post the release notes to #eng' — the plugin routes the call through the shared Arcade gateway automatically.
Introduces the Arcade Plugin, installable in Claude Code, Claude Desktop, Cursor, VS Code, GitHub Copilot CLI, and Codex, connecting agents to a single shared Arcade gateway without MCP gateway configuration.︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄ Not checked yet Adds Tool Recommendation capability that keeps a catalog of over 1K agent-optimized tools out of the model context window context window The maximum amount of text, counted in tokens, a model can take in and produce in one request. Everything the model knows about the task has to fit inside it. , enabling one gateway to serve all users dynamically. Not checked yet Ships Skills and lifecycle hooks inside the plugin that teach agents when to invoke Arcade tools and how to authorize services like Gmail or Slack from inside the agent client. Not checked yet
3 details in Everything
Arcade adds workload-identity support, single Postgres URL config, resource-quota safety gating, a Helm values schema, and credential redaction in tracing spans.
Adds sslmode: verify-full Postgres connection verification against the system trust store, plus support for a single Postgres connection URL as an alternative to individual config values. Not checked yet Adds deployments.resourceQuota.enabled and deployments.sizing.* Helm chart values to control resource quotas and pod sizing. Not checked yet Adds workload-identity support for the engine pod (EKS EKS Amazon's managed Kubernetes service, run by AWS, that handles control-plane provisioning so teams can deploy containerized workloads without operating the underlying cluster infrastructure themselves. IRSA IRSA An AWS mechanism that lets a Kubernetes pod assume an IAM role via a service account token, so cyber tools running in EKS can call AWS APIs without storing long-lived credentials. , Azure Workload Identity, GKE GKE Google Kubernetes Engine, Google's managed Kubernetes service, which lets cyber tools run containerized workloads without managing the underlying cluster infrastructure. Workload Identity, or a mounted GCP key) via service account annotations, pod labels/annotations, and extra volumes/mounts. Not checked yet Adds a values schema to the Helm chart that documents and type-validates every chart value. Not checked yet Worker pod creation now waits for an applied resource quota before proceeding, acting as a fail-closed safety gate. Not checked yet
+10 more
show less
Tool-recommendation requests run concurrently instead of serially, reducing latency. Not checked yet New gateways default to tool recommendation automatically when the org has it enabled. Not checked yet Versionless tool names now resolve to the actual newest version. Not checked yet Tool listings return an error instead of silently showing no tools when the tool store cannot be read. Not checked yet Optional tool inputs: an explicit null now applies the field default instead of failing the call. Not checked yet MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. gateway explorer UI follows system light/dark theme with a toggle and shows the Arcade favicon. Not checked yet GitHub review-comment listing response adds outdated status, line numbers, diff side, and reply-to id fields, and now returns the full set on large pull requests. Not checked yet Notion tables in page content now render as Markdown tables. Not checked yet Redis spans no longer include raw command text, preventing live API key exposure to anyone with tracing-backend access. Not checked yet Engine HTTP spans no longer include credential header values (Authorization , Cookie , API-key, tool-token). Not checked yet
15 details in Everything
└──▷ BREAKING ON UPGRADE
! Google Calendar 5.0.0 is a breaking rebuild for unpinned callers — review Upgrade notes before upgrading; it also drops the delete an event and 'find free time slots across attendees' tools (those remain available on 4.2.4).! format=oxp on tool listings is removed.! The engine image will not start on the old database schema; the migration runs automatically on helm upgrade but is required before the new engine image starts.! Two MCP Apps feature flags are removed; the features are now always on.! The engine no longer uses a curl init container to wait for worker readiness.! A busybox-based init container that copied env vars into a redundant file is removed.! Several images now pull from a different registry — review and update registry egress rules before upgrading.! A small set of toolkit parameters that previously silently ignored null now reject it; callers must omit those parameters instead of passing null.! The GitHub review-comment listing response shape changes: the truncated field is removed.! The default curl support image chart value is deprecated and will be removed in a future chart version.
Arcade now publishes an API — 59 endpoints across 8 areas: Admin, Tools, Gateways, …︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄
Admin (21 endpoints) — Project administration: auth providers, secrets, user connections, and session verification settings.︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄ Not checked yet Tools (12 endpoints) — Discover tools, inspect their authorization requirements, and execute them by name. Not checked yet Gateways (7 endpoints) — MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. gateways — federate tools from multiple MCP servers behind a single URL. Not checked yet Hooks (6 endpoints) — Runtime hooks that apply your own policy (DLP DLP Data Loss Prevention: a category of controls that inspect and block sensitive data from leaving an organisation via email, web, or removable media, giving cyber tools a policy enforcement layer over data in motion and at rest. , approval, redaction) to tool calls. Not checked yet Plugins (5 endpoints) — Plugins installed into the Engine and their configuration. Not checked yet
+3 more
show less
Authorization (4 endpoints) — Start and check the user authorization flows that grant a tool the provider scopes it needs. Not checked yet Operations (3 endpoints) — Operational reads: tool execution history, counts, and scheduled tools. Not checked yet Llm (1 endpoint) — OpenAI-compatible chat completions with tool calling tool calling A model's ability to answer with a structured request to run a named function or API, which the application executes and returns. It is how agents take actions. handled by the Engine. Not checked yet
8 details in Everything
Arcade launches Tool Recommendation, dynamically fetching only needed tools via Arcade_SelectTools and Arcade_UseTool MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. built-ins, cutting input tokens by up to 91%.︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄
Adds Arcade_ListApps, Arcade_SelectTools, and Arcade_UseTool as built-in tools exposed through the Arcade gateway, replacing full catalog loading with on-demand tool retrieval over MCP.︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄ Not checked yet Adds Tool Recommendation, a new capability that uses purpose-built embedding embedding Lists of numbers a model produces to represent the meaning of text, images or code, so similar items sit close together. They power semantic search and RAG. and ranking models to match an agent's plain-language task description against the full tool catalog and return only the relevant tool schemas. Not checked yet Enables Tool Recommendation per Arcade gateway with no SDK changes, framework-specific integration, or agent-side rebuilds — works with any MCP client or framework. Not checked yet Applies pre- and post-tool-call policy enforcement hooks, PII PII Personally Identifiable Information — data that can identify a specific individual, such as names, email addresses, or SSNs. Cyber tools flag or redact it to limit exposure and meet privacy regulations. screening, and DLP DLP Data Loss Prevention: a category of controls that inspect and block sensitive data from leaving an organisation via email, web, or removable media, giving cyber tools a policy enforcement layer over data in motion and at rest. protection to every recommendation and tool call, with audit trails exported via OpenTelemetry OpenTelemetry An open standard and set of tools for collecting traces, metrics and logs from applications. It lets services export observability data without tying instrumentation to one monitoring vendor. . Not checked yet Available in Arcade Cloud now; also deployable in customer cloud, on-premises, and air-gapped environments connected to existing identity providers identity providers A service that authenticates users and supplies identity information to other applications. It is the sign-in authority behind many single sign-on systems. and policy engines. Not checked yet
5 details in Everything
Arcade adds Nimble as a partner MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. server integration for real-time web search.
Adds Nimble to the list of supported Partner MCP Servers for real-time web data retrieval. Not checked yet
1 detail in Everything
Arcade adds resolve_comment Linear tool, configurable MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. /list page size, project-scoped execution API, Gmail unsubscribe tool, and UI catalog/search upgrades
[MCP] Add secure Gmail unsubscribe tool with various flow outcomes and enhanced validation Not checked yet [MCP] Add resolve_comment tool to Linear toolkit for managing comment resolution state. Not checked yet Platform and Engine Make MCP tools /list page size configurable. Not checked yet Platform and Engine Adds project-scoped tool execution list and detail API for enhanced visibility of tool-execution history. Not checked yet Platform and Engine Migrates the dashboard Tools page to a catalog-driven, server-side search with lazy loading. Not checked yet
+1 more
show less
Platform and Engine Implement undo toast notification for destructive picker edits in UI. Not checked yet
6 details in Everything
Arcade adds stateless MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. paths, Gmail label lifecycle tools, and structured 5xx logging.
Adds tools/call and /list as stateless MCP endpoints on the real catalog Not checked yet Supports per-request protocol-version dispatch on /mcp Not checked yet Implements MRTR url-mode elicitation for tool OAuth challenges on the stateless MCP path Not checked yet Adds Gmail tools for label lifecycle management and conversation trash status handling Not checked yet Adds Arcade.ListApps builtin to prefix apps into MCP prompt Not checked yet
+2 more
show less
Adds read tool to retrieve Linear project status updates with pagination Not checked yet Adds structured logging for unhandled 5xx exceptions with route and type information Not checked yet
7 details in Everything
Arcade adds Microsoft Power BI and Postman MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. toolkits, configurable chat-model selector, and dedicated MCP gateway pages.
Adds microsoft_powerbi toolkit for Power BI integrations in Arcade MCP with 4 new management tools. Not checked yet Adds Postman API integration with 30 tools and API-key authentication. Not checked yet Introduces a configurable chat-model selector allowing dynamic selection based on a curated catalog. Not checked yet Introduced dedicated create/edit pages for MCP gateways with new routing and tool selection. Not checked yet Enables default RBAC RBAC Role-based access control, where permissions are assigned to roles such as reader or administrator and users receive those roles. It makes access easier to manage than individual permission lists. for new Cloud signups and surfaces can_invite on org endpoints. Not checked yet
5 details in Everything
Arcade adds Insightly and Fireflies MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. toolkits, a new server picker with custom hooks, and initial RBAC RBAC Role-based access control, where permissions are assigned to roles such as reader or administrator and users receive those roles. It makes access easier to manage than individual permission lists. groundwork.
Insightly CRM toolkit with 29 tools for sales/account management is now available as an MCP server integration Not checked yet Fireflies.ai meeting-intelligence toolkit with 17 functionalities is now available as an MCP server integration Not checked yet New MCP server picker types and a custom hook enable selection/search of MCP servers directly in the UI Not checked yet Preparations for RBAC functionality have been added to the platform, enabling future role-based access control role-based access control Role-based access control, where permissions are assigned to roles such as reader or administrator and users receive those roles. It makes access easier to manage than individual permission lists. Not checked yet
4 details in Everything
Arcade adds Google Workspace Drive tools, OneDrive toolkit, Telegram interactive messaging, and CLI operator authorization.
Adds search_directory_people tool for resolving coworkers’ emails in Google Workspace Not checked yet Adds search_drive tool to allow searching the user’s entire Google Drive Not checked yet Adds check_file_access pre-flight probe and file_types filter for Google Drive access improvements Not checked yet Adds interactive messaging with buttons and response handling to the Telegram worker toolkit Not checked yet Adds Microsoft OneDrive toolkit with 16 tools for file management Not checked yet
+2 more
show less
Enables CLI command for operator authorization in RBAC RBAC Role-based access control, where permissions are assigned to roles such as reader or administrator and users receive those roles. It makes access easier to manage than individual permission lists. Not checked yet Introduces O(1) predicate for single-name lookups in MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. Not checked yet
7 details in Everything
Arcade adds pipeline exit archival, VS Code chat host support, X toolkit rebuild, and Agent Capability Gap MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. builtin tool
Adds pipeline_exit functionality to archive applications in the Ashby toolkit via Arcade MCP Servers. Not checked yet Adds vs_code_chat_host support for attachment capabilities in agent mode. Not checked yet Provides a ground-up rebuild of the X toolkit within Arcade MCP Servers. Not checked yet Overhauls Google Slides integration with new improvements and fixes. Not checked yet Adds deep-validation for OIDC OIDC OpenID Connect, a sign-in protocol built on OAuth 2.0. An application uses identity information from a trusted provider instead of maintaining a separate password for each user. user source issuers and improves coordinator error detail visibility in Platform and Engine. Not checked yet
+1 more
show less
Introduces Agent Capability Gap self-report MCP builtin tool with master switch and LLM LLM A large language model, a neural network trained on large amounts of text to recognize patterns and generate language. Applications use it for tasks such as answering, summarizing and writing code. instruction integration. Not checked yet
6 details in Everything
Arcade adds 13 new MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. toolkits and expands Google Sheets, Linear, Gmail, and Discord capabilities.
Adds Freshdesk customer-support toolkit with API-key authentication Not checked yet Adds Ashby recruiting/ATS toolkit for core recruiting operations Not checked yet Adds Vercel toolkit for managing accounts via the REST API REST API A web API design style where clients interact with resources over HTTP using standard methods like GET, POST, and DELETE, giving cyber tools a predictable, language-agnostic way to integrate with external services. Not checked yet Adds GetSpreadsheetEditHistory tool to google_sheets toolkit Not checked yet Adds Datadog observability toolkit for log and trace investigation Not checked yet
+9 more
show less
Adds project -milestone assignment to Linear's create and update issue tools Not checked yet Adds inspect_spreadsheet tool and removes legacy read tools in Google Sheets integration Not checked yet Adds first-party Discord toolkit for bot functionality in worker applications Not checked yet Returns deep links to the booking page for Google Flights Not checked yet Adds support for attaching local files to emails in Gmail via new attachments parameter Not checked yet Adds Fly.io toolkit with tools for managing apps and infrastructure Not checked yet Adds read-only Apollo.io sales-intelligence toolkit Not checked yet Adds Customer.io MCP toolkit for customer engagement Not checked yet Adds empty-state page with hero and functional Create button for MCP gateways Not checked yet
14 details in Everything
└──▷ BREAKING ON UPGRADE
! Removes legacy read tools in the worker's Google Sheets integration (replaced by inspect_spreadsheet)
Arcade adds GitHub SearchCode MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. server, configurable OIDC OIDC OpenID Connect, a sign-in protocol built on OAuth 2.0. An application uses identity information from a trusted provider instead of maintaining a separate password for each user. scopes, User Source audit logging, and new toolkits for Outlook, Resend, Google Sheets, Calendar, and Google Fligh
Adds the Github_SearchCode MCP server for cross-organizational code search Not checked yet Adds create_or_edit_spreadsheet tool for batch operations on Google Sheets Not checked yet Rebuilds the arcade_google_flights toolkit to support full trip-planning lifecycle Not checked yet Adds ChatGPT as a supported MCP client in the mcp-gateway-connect feature Not checked yet Adds move email to folder tools in the Microsoft Outlook Mail toolkit Not checked yet
+4 more
show less
Adds Resend transactional email toolkit with API-key authentication for email operations Not checked yet Adds support for recurring events in Google Calendar API integration Not checked yet Makes OIDC scopes configurable per User Source, enhancing integration with Microsoft Entra ID Entra ID Microsoft's cloud-based identity and access management service, formerly Azure Active Directory, used by cyber tools to authenticate users and enforce access policies in Microsoft 365 and Azure environments. Not checked yet Adds User Source CRUD CRUD The four basic database operations — Create, Read, Update, Delete — used as shorthand for any system that manages persistent data records, giving cyber tools a standard model for interacting with stored information. audit logging and emission framework in the coordinator Not checked yet
9 details in Everything
Arcade introduces static-user-source gateway login and reshaped GitHub MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. server PR objects.
Reshapes PR objects to a nested structure for the GitHub MCP server integration Not checked yet
1 detail in Everything
Adds default-on Gmail filter for MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. Servers and configurable tool-call timeouts for self-hosted Arcade deployments.
Add default-on filter to exclude automated noise from Gmail search/list results in Arcade MCP Servers Not checked yet
1 detail in Everything
Arcade adds domain modules to its first-party MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. server and OAuth bearer path handling for user_source gateways
Major overhaul of the first-party MCP with new domain modules and tools Not checked yet
1 detail in Everything
Arcade 2026-05-01 adds Linear milestone management, Telegram toolkit, and partner badge.
Linear milestone management tools (queries/mutations) for project planning workflows Not checked yet Telegram toolkit with core messaging tools using Telegram Bot API Not checked yet Partner badge and Tavily as first partner MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. Not checked yet
3 details in Everything
Arcade 2026-04-24 adds ?connect param to MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. gateways list, Google Drive permissions tools, and CC support in draft replies.
Add a connect search param to the MCP gateways list route Not checked yet Revamped GoogleDocs.EditDocument with enhanced editing capabilities Not checked yet Add CC support to draft replies and improve handling of email threads Not checked yet Adds permission management tools to the GoogleDrive MCP server Not checked yet
4 details in Everything
Arcade adds configurable MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. gateway header passthrough and Linear MCP unassign capability
Adds gateway.headers configuration key to .arcade/mcp_servers.yaml to control HTTP header passthrough for MCP gateways. Not checked yet Adds the unassign_issue tool to the Linear MCP server, allowing tools to remove assignees from issues. Not checked yet Enhances the Gmail MCP server's structured output structured output Forcing a model's response to match a given format or JSON schema, so software can parse it reliably instead of hoping the text is well formed. to explicitly include thread participant emails in the participants field. Not checked yet
3 details in Everything
Arcade adds AG2 framework support, a PostHog toolkit with 41 tools, and Claude Code gateway auth
Adds AG2 as a supported agent framework in the docs with a new walkthrough. Not checked yet Adds Arcade MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. Servers formatting support to GoogleDocs.CreateDocumentFromText to interpret input as Markdown. Not checked yet Adds Arcade Auth support to Claude Code gateway connect option. Not checked yet
3 details in Everything
Arcade adds TelemetryPassbackMiddleware and full MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. Resources support with enhanced dashboard auth.
Adds TelemetryPassbackMiddleware for serverExecutionTelemetry capability in arcade-mcp Not checked yet Adds full support for MCP Resources, enhancing resource registration and JSON Schema generation in arcade-mcp-server Not checked yet Upgrades dashboard authentication experience with a dedicated error page for OAuth failures Not checked yet
3 details in Everything
└──▷ BREAKING ON UPGRADE
! Outlook Calendar, Outlook Mail, and SharePoint toolkits are renamed to microsoft_* prefix; old packages are deprecated
Arcade MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. adds Figma raw data inspection and a dedicated Microsoft Outlook/Calendar toolkit
Adds include_raw_data option to Figma_GetFileNodes for enhanced node detail inspection Not checked yet Releases a separate Microsoft toolkit for Outlook Mail & Calendar functionalities Not checked yet
2 details in Everything
└──▷ BREAKING ON UPGRADE
! The outlook and sharepoint toolkits are renamed to follow the standardized microsoft_* prefix; old packages are DEPRECATED
Arcade adds non-interactive MCPApp package generation via arcade new flag and multi-select filters in audit logs API.
Adds --full flag to arcade new <name> for non-interactive MCPApp package generation Not checked yet Allows selecting multiple filters at once in the audit logs API Not checked yet
2 details in Everything
Arcade adds YugabyteDB, MongoDB, ClickHouse, and Postgres toolkits with ToolMetadata support.
Add YugabyteDB toolkit Not checked yet Add MongoDB toolkit Not checked yet Add Clickhouse toolkit Not checked yet Add Postgres toolkit Not checked yet
4 details in Everything
Arcade adds Batch Tool Reconcile and grouped toolkit overview to the Dashboard.
Arcade adds Ed25519 Ed25519 A fast elliptic-curve signature scheme with small keys, the default for new SSH keys and common in code signing. It is not quantum-safe. support to MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. servers and dashboard expiring API key management.
arcade-mcp supports the Ed25519 algorithm for server authentication Not checked yet Dashboard now includes UI controls to manage expiring API keys Not checked yet
2 details in Everything
Arcade introduces OAuth for MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. Gateways, per-project secret sharing, member listing APIs, and new Google/ PagerDuty toolkits
Adds OAuth support for MCP Gateways to enable federated authentication flows Not checked yet Allows the same secret to be stored in multiple projects via shared reference Not checked yet Supports listing members and invited users for a given project via API Not checked yet Adds optional image_url and hex_color fields for organization and project resources Not checked yet Enables [Google Contacts] phone number support in the MCP App Not checked yet
+1 more
show less
Converts [Gmail] integration to an optimized MCP App with improved performance Not checked yet
6 details in Everything
Arcade 2025-12-12 adds OAuth for MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. servers, multi-org CLI support, and Google Drive/contacts toolkits.
Adds OAuth authentication configuration for arcade-mcp servers Not checked yet Enables running multiple uvicorn workers for arcade-mcp servers Not checked yet Includes type annotations in the arcade_mcp_server module Not checked yet Supports specifying multiple organizations and projects via Arcade CLI flags Not checked yet Adds phone number support in Google contacts toolkits Not checked yet
+1 more
show less
Enables file download and upload operations to Google Drive Not checked yet
6 details in Everything
Arcade adds GitHub/Ashby/Linear toolkits, email invites, CLI endpoints, and startup warnings for missing secrets.
Add tools for project management, pull request, and projects to the Github MCP Server MCP Server A process that exposes tools, resources, and prompts over Anthropic's Model Context Protocol, letting an AI agent call external APIs or local services through a standardised interface. Not checked yet Add Optimized Linear toolkit Not checked yet Add Optimized Ashby toolkit Not checked yet
3 details in Everything
Arcade adds GitHub MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. support for repos/PRs and email-based user invitations.
MCP Servers updated to support GitHub projects, issues, and pull requests Not checked yet Platform and Engine enables inviting users to projects by email Not checked yet
2 details in Everything
Adds Customer.io and Intercom MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. servers; updates arcade deploy to support them
Adds Customer.io Starter MCP Server to Arcade Not checked yet Adds Intercom Starter MCP Server to Arcade Not checked yet Updates arcade deploy command to support MCP Servers built with arcade-mcp Not checked yet
3 details in Everything
Adds Mailchimp market toolkit, Hubspot Marketing & CRM toolkit enhancements, and optional OAuth header customization for upstream MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. Servers
Added Mailchimp market toolkit as a new integration. Not checked yet Enhanced Hubspot Marketing & CRM toolkit capabilities. Not checked yet Added optional customization of OAuth request header format for upstream MCP Servers. Not checked yet
3 details in Everything
Arcade 2025-10-31 adds seven new starter toolkits and the arcade deploy CLI command.
arcade deploy subcommand available for platform deployment Not checked yet New HubSpot Marketing & CRM starter toolkit Not checked yet New Exa.ai Starter MCP Server MCP Server A process that exposes tools, resources, and prompts over Anthropic's Model Context Protocol, letting an AI agent call external APIs or local services through a standardised interface. integration Not checked yet New Asana starter toolkit Not checked yet New Github starter toolkit Not checked yet
+3 more
show less
New Pylon Starter Toolkit Not checked yet New Posthog Starter Toolkit Not checked yet New Clickup Starter Toolkit Not checked yet
8 details in Everything
Arcade adds five new Starter Toolkits and OAuth/config controls for MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. Servers in the Dashboard.
Dashboard now accepts redirect_uri configuration for MCP Servers Not checked yet Dashboard exposes OAuth fields configuration for MCP Servers Not checked yet Adds Ticktick Starter Toolkit to Arcade Not checked yet Adds Weaviate Starter Toolkit to Arcade Not checked yet Adds Vercel Starter Toolkit to Arcade Not checked yet
+1 more
show less
Adds Datadog Starter Toolkit to Arcade Not checked yet
6 details in Everything
Arcade 2025-10-17 adds Projects, MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. Gateways GA, arcade-mcp GA, BrightData/Figma/AirTable/Miro/PagerDuty/Cursor agents toolkits, and OAuth support for MCP Servers.
Adds Projects as a first-class resource to organize MCP Servers, tools, and secrets Not checked yet MCP Gateways are now generally available for federating multiple MCP Servers Not checked yet arcade-mcp toolkit is now generally available, superseding Arcade TDK Not checked yet Adds BrightData Starter MCP Server MCP Server A process that exposes tools, resources, and prompts over Anthropic's Model Context Protocol, letting an AI agent call external APIs or local services through a standardised interface. integration Not checked yet Adds Figma Starter MCP Server integration Not checked yet
+7 more
show less
Adds Freshservice Starter MCP Server integration Not checked yet Adds Cursor Agents Starter MCP Server integration Not checked yet Adds AirTable starter MCP Server integration Not checked yet Adds Miro Starter MCP Server integration Not checked yet Adds PagerDuty Starter MCP Server integration Not checked yet Enables arcade deploy for MCP Servers built with arcade-mcp Not checked yet Dashboard now allows configuring OAuth on MCP Servers Not checked yet
12 details in Everything
Arcade adds Trello, Calendly, SquareUp, Xero toolkits; reference area for MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. docs
Trello toolkit now available in Arcade Not checked yet Calendly starter toolkit added Not checked yet SquareUp toolkit added Not checked yet Xero API Starter MCP server included Not checked yet Reference area added to arcade-mcp documentation (PR #488) Not checked yet
5 details in Everything
Arcade adds Box.com and Stripe Starter MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. Servers
Adds Box.com Starter MCP Server for cloud storage automation. Not checked yet Adds Stripe Starter MCP Server for payment processing integration. Not checked yet
2 details in Everything
Arcade introduces unoptimized toolkits, a Slack-started MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. server, and WhoAmI tools for Confluence.
Advanced error handling added to Google, Microsoft, Slack, and Asana MCP Servers Not checked yet
1 detail in Everything
Adds MongoDB and HubSpot MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. servers; removes noisy messages in ClickUp Evals
Adds the MongoDB analytics MCP Server MCP Server A process that exposes tools, resources, and prompts over Anthropic's Model Context Protocol, letting an AI agent call external APIs or local services through a standardised interface. for toolkits (PR #548) Not checked yet Adds HubSpot MCP server enhancements to the MCP Hub (PR #441) Not checked yet Removes no content additional messages in ClickUp Evals Not checked yet
3 details in Everything
Arcade introduces WhoAmI tool across Jira, Google, Clickup, Slack, and Zendesk MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. Servers.
Adds Imgflip MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. and Google Docs editing toolkits plus new dashboard sidebar for Arcade
Adds the Imgflip MCP Server toolkit providing tools for creating memes via MCP protocol. Not checked yet Adds the Edit Google Document Tool to the available toolkits list. Not checked yet Introduces a new sidebar and user-verification page in the Dashboard UI. Not checked yet
3 details in Everything
Arcade introduces Twitter reply capability, Jira sprint management tools, and engine authentication improvements in v2025-08-22.
Adds Reply to Tweet tool to the X (Twitter) Toolkit for automated tweet responses Not checked yet Adds Add To Sprint and Remove from Sprint tools to the Jira Toolkit for sprint workflow automation Not checked yet Enables authentication between Arcade Engine and Coordinator via key exchange Not checked yet
3 details in Everything
Arcade 2025-08-15 adds Sharepoint, Google Slides toolkits and custom OAuth support; enforces unique callback URLs.
Sharepoint Toolkit added to Arcade for Microsoft SharePoint integrations Not checked yet Google Slides Toolkit added to Arcade for Google Slides integrations Not checked yet Commenting on Google Docs capability added to existing toolkits Not checked yet Custom OAuth providers now require a unique callback URL Not checked yet
4 details in Everything
Arcade adds Clickhouse toolkit, Google Drive search, GPT-5 support, and per-app redirect URIs
Introduces the Clickhouse Toolkit for connecting to ClickHouse databases Not checked yet Adds search capability within Google Drive integrations Not checked yet Enables using GPT-5 models in tool invocations Not checked yet Provides per-app redirect URI information for OAuth configurations Not checked yet
4 details in Everything
Arcade adds Microsoft Teams and Google Sheets MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. servers, Jira toolkit tools, Salesforce config via secrets, and Confluence multi-cloud CLI support.
Adds Microsoft Teams MCP Server to enable Microsoft Teams integration Not checked yet Adds Jira Toolkit with new List Sprints & Boards tools for Jira automation Not checked yet Adds pagination support to the GetSpreadsheet tool in the Google Sheets MCP Server MCP Server A process that exposes tools, resources, and prompts over Anthropic's Model Context Protocol, letting an AI agent call external APIs or local services through a standardised interface. Not checked yet Enables configuration of subdomain and max concurrency for the Salesforce MCP Server via secrets Not checked yet Extends Confluence MCP Server support to Atlassian multi-cloud CLI and TDK Not checked yet
5 details in Everything
Arcade adds value schema support for tool outputs, multiple Jira cloud support, and self-service plan selection
ToolDefinition supports using Tool Output within the ValueSchema, enabling dynamic value generation based on prior tool execution results Not checked yet Jira Toolkit now supports configuration against multiple Atlassian Cloud instances simultaneously Not checked yet Arcade Cloud enables self-service plan selection and payment directly through the dashboard Not checked yet
3 details in Everything
└──▷ BREAKING ON UPGRADE
! MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. Servers previously hosted in github.com/ArcadeAI/arcade-ai are now closed-source; source code is available only upon request for paying customers
Arcade v2.0 adds Linear and Zendesk toolkits, CLI callback_host flag, and Dashboard UI for security settings.
Adds linear toolkit for Arcade Not checked yet Adds zendesk toolkit for Arcade Not checked yet Adds callback_host configuration option for the arcade login command to specify a custom callback host Not checked yet Adds an OAuth endpoint for user confirmation in the Engine Not checked yet Adds API endpoints for scheduled subscriptions in the Platform Not checked yet
+1 more
show less
Adds user id and providers filters in the Connected Accounts section of the Dashboard Not checked yet
6 details in Everything
└──▷ BREAKING ON UPGRADE
! Self-hosted upgrades to Engine v2.0.0+ force the default user verifier, removing the ability to skip user verification.
Arcade 2025-07-11 splits toolkits into separate MCP MCP Model Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. servers and adds CLI --debug flag
Adds --debug flag to CLI commands for verbose output Not checked yet Splits combined Google, Microsoft, and other Toolkits into separate MCP Servers Not checked yet
2 details in Everything
Concurrent auth deduplication, secret description editing, and parameter reset preservation in Arcade Dashboard
Enables concurrent authentication requests for the same user/scopes to share a single flow and URL Not checked yet Adds ability to edit the description of a secret in the Dashboard Not checked yet Ensures tools are preserved when resetting parameters in the Dashboard Not checked yet
3 details in Everything