Jentic

A dispatcher: four tools reaching thousands of third-party APIs, since execute takes the operation as an argument.

Add to Agentman Studio

Opens Agent Studio, where connecting is one click. The connector URL below works in any MCP client.

Paste it into any MCP client. Setup docs

Resources

What the Jentic connector does

The Jentic MCP server is a dispatcher: rather than exposing one tool per action, it exposes four tools that reach a catalogue of thousands of third-party APIs. You search for an operation in plain English, load its contract, then run it — with Jentic injecting the stored credential. The tool count bounds nothing, because the destination is an argument.

Verified connector

Listed by Anthropic as a partner connector in its Connectors Directory.

Connection checked by Agentman on .

Anthropic states this reflects the level of review a connector received, not a security audit.

Jentic tools (4)

  • list_credentialsRead-only

    Use this tool to list all API credentials the authenticated agent has access to. Use this if a user asks "What APIs can I use?", "What credentials do I have access to?", "What apps are available to me?", "What APIs are available to me?" This is useful for debugging or to understand what API credentials are available to the agent. DEPRECATION NOTICE: This tool is being deprecated on 20 September 2026. Execution is moving to self-hosted Jentic One. It still works for now but will stop working on that date. See https://jentic.com/jentic-one, or fetch the 'skill://onboard-jentic-one' MCP resource for a step-by-step install and onboarding guide. Returns: dict: Response with success status and list of credentials { "result": { "success": True, "credentials": [c.model_dump(exclude_none=False) for c in credentials], } } On error: { "result": { "success": False, "message": "Failed to load credentials", } }

  • search_apisRead-only

    Use this tool to explore for available actions or information based on what the user wants to do (e.g., 'find Discord servers', 'send a message'). Use this to discover what APIs and operations exist in the Jentic catalogue. Execution is moving to self-hosted Jentic One: to run a discovered operation, install Jentic One (https://jentic.com/jentic-one) and use its search -> inspect -> execute flow through the local broker.

  • load_execution_infoRead-only

    Use this tool to get the full execution contract for a specific operation (its inputs, outputs, and authentication requirements). This tool must use the operation UUIDs returned from the search_apis tool. DEPRECATION NOTICE: This tool is being deprecated on 20 September 2026. Execution is moving to self-hosted Jentic One. It still works for now but will stop working on that date. To load an operation's contract and run it, install Jentic One (https://jentic.com/jentic-one) and use its `jentic inspect` / `jentic execute` flow. For a step-by-step install and onboarding guide, fetch the 'skill://onboard-jentic-one' MCP resource. Example operation UUIDs (returned from search_apis tool): NewsAPI.org GET /everything: op_ba86fdce1bade1b7 NewsAPI.org GET /top-headlines: op_8ae297966b7f6da3 Finnhub.io GET /quote: op_f97a5db5e23f7b5a Coincap.io GET /v3/assets/"slug": op_e4a7f50d57fee805

  • executeDestructive

    Use this tool to execute the chosen action for the user using the provided details. Always include `inputs` even if there are none. To reduce context, use `include_paths` to return only the fields you need from the output. DEPRECATION NOTICE: This tool is being deprecated on 20 September 2026. Execution is moving to self-hosted Jentic One. It still works for now but will stop working on that date. To keep running operations, install Jentic One (https://jentic.com/jentic-one) and execute through its local broker. For a step-by-step install and onboarding guide, fetch the 'skill://onboard-jentic-one' MCP resource. Example operation UUIDs (from search_apis / load_execution_info): NewsAPI.org GET /everything: op_ba86fdce1bade1b7 NewsAPI.org GET /top-headlines: op_8ae297966b7f6da3 Finnhub.io GET /quote: op_f97a5db5e23f7b5a Coincap.io GET /v3/assets/"slug": op_e4a7f50d57fee805

Read from the server on 2026-08-22, including each tool's own safety annotations.

Limits

  • The tool count bounds nothing. Four tools reach a catalogue we measured at 6,323 API documents and 1,999 named workflows. Any assessment of this connector that stops at the tool list has measured the dispatcher and not the dispatch.
  • Jentic's headline API count is not reproducible from its own repository. The directory description and Jentic's site say 10,000+; the repository's manifest carried 6,323 entries on 2026-08-22, and the only 10,000+ in it is a badge image's alt text.
  • No application scopes exist. openid, email and profile are the complete set, on both the resource descriptor and the authorization server. Nothing at the OAuth grant separates reading from writing, or one vendor from another.
  • The access boundary is invisible at consent time. It lives in an MCP Configuration set on app.jentic.com, chosen before you connect and never re-shown by Claude.
  • Governance controls are a paid tier. Jentic's pricing page places "Governance and integration controls" in Enterprise Edition; free Core Edition lists none.
  • Execution is deprecated on 2026-09-20. execute, load_execution_info and list_credentials all say so in their own descriptions. Search continues.
  • The catalogue is partly AI-generated and self-labelled ALPHA. The jentic-public-apis README states Jentic generated "thousands" of Arazzo workflows with AI and generates OpenAPI specifications "where no structured documentation previously existed", and the repository says it "is currently in ALPHA". An operation contract is a description of an API, not the API — a wrong contract means a wrong call.
  • Anyone can open a catalogue submission, and CI drafts it automatically. The repository's import-openapi.yaml GitHub Action triggers on any newly opened issue whose body contains a spec URL, fetches it, generates the catalogue entry and opens a pull request against main. There is no auto-merge: the contributing guide states a maintainer reviews and merges. So the single gate on the reachable operation set is human PR review.
  • We could not confirm the hosted service serves this exact branch. Nothing in the repository states whether api.jentic.com reads main directly or a separately curated snapshot, and establishing it would require authenticated access we did not use.
  • Free-tier execution logs train Jentic's models. Its privacy policy states that as a condition of the free tier it may analyse "agent activity traces and execution logs" to "develop and train our own custom AI models", using anonymised and aggregated content.
  • You are the data controller. The same policy states that when your agents process personal data through third-party integrations "you remain the data controller", including for international transfers.
  • The directory's auth posture is wrong. auth_posture: "auth_required" and a recorded 401 do not describe connecting: anonymous initialize and tools/list both returned 200.
  • One tool is missing from the directory and from the vendor's tool table. list_credentials is live but appears in neither, and it enumerates your stored credentials.
  • Prompts drift. The server declares a prompts capability with listChanged: true, and prompts/list returns an empty array.
  • We did not call a single tool. Every call requires OAuth, and we did not authenticate. All behavioural claims above come from schemas, annotations, server-served documents and Jentic's own documentation — never from execution.

Frequently asked questions

How many things can the Jentic MCP server actually do?

Four tool names, but the reachable action set is the whole Jentic catalogue. The execute tool takes an operation UUID as an argument, so the operation is data rather than a tool name. We counted 6,323 API documents and 1,999 named workflows in Jentic's open catalogue on 2026-08-22. Counting four tools measures nothing.

What OAuth permissions does the Jentic MCP server request?

Only openid, email and profile. The server's own RFC 9728 descriptor and its authorization server both advertised exactly those three on 2026-08-22, and all three are OpenID identity claims. No scope names an API, a vendor, or a read-versus-write distinction, so the consent screen cannot separate looking something up from sending something.

Can the Jentic connector send email, post messages or move money?

Yes, where you have configured a credential for a service that does. Jentic's catalogue carries SendGrid's send-mail operation, Twilio's send-SMS and place-call operations, Discord's post-message operation and Stripe. Nothing in the four tool names says any of this, because the destination is an argument rather than a verb.

Why does the Jentic connector list its tools without asking me to sign in?

Discovery is open and execution is closed. An anonymous handshake on 2026-08-22 returned the full tool list, the resource list and the initialize capabilities with no credential. Every tools/call, including read-only ones, returned 401 with a www-authenticate header. So the listing is public but nothing runs without OAuth.

Which APIs can the Jentic connector reach in my account?

Whatever is in the MCP Configuration set you built at app.jentic.com, which Claude never shows you. Jentic's Claude Desktop guide makes creating that set a prerequisite before connecting, and you add credentials to it there. The boundary is therefore chosen in Jentic's dashboard, separately from and before Claude's consent screen.

Is the Jentic execute tool being discontinued?

Yes, on 20 September 2026. Three of the four tools carry a deprecation notice in their own live descriptions saying execution is moving to self-hosted Jentic One. Jentic's Claude Code plugin repository describes the hosted server as a search tool and says the load and execute path is deprecated. Search is not affected.

Does Jentic see the data my agent sends to third-party APIs?

Yes, it passes through Jentic's hosted service. Jentic's privacy policy states it processes your prompts, instructions and the responses generated, and that on the free tier it may use anonymised execution logs and agent activity traces to train its own models. It also states you remain the data controller for anything your agents process.

Who decides which APIs are in the Jentic catalogue?

Jentic and community contributors, gated by maintainer pull-request review. The catalogue is the open jentic-public-apis repository, where a GitHub Action turns any issue carrying a spec URL into a draft pull request. Its README states specifications were generated with AI where none existed, and the repository labels itself ALPHA.

Sources

More Productivity connectors

  • 6sense

    Ask an AI agent about 6sense account insights, buying signals and keyword trends.

  • Adobe Workfront

    Find, create, update and delete Workfront work items from an AI agent.

Browse all MCP connectorsClaude and ChatGPT setup guideBrowse AI skills

Product names and logos belong to their owners. Agentman is not affiliated with a connector's maker unless the page says the connector is built and run by Agentman.

Use it in an agent

Put Jentic to work.

Connect Jentic once and your agents call these tools on their own: on a schedule, in a workflow, with nobody at the keyboard.

Open in Agentman Studio