Universal Connector MCP

Install

Works with Claude Code, Codex, VS Code, Cursor, Windsurf, Claude Desktop. Read the repository before you give any tool access to your data.

npx agentmarket install connector/teodormcp-universal-connector-mcp

About

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this server can do and what it asks for, from its tool list and install config. Checked Oct 25, 2026.

Tools

5 read · 3 write · 1 delete · 1 not yet classified

Secrets

1 (API keys, tokens or passwords)

Install scripts

Not scanned yet

Runs as

Python package

Tools (10)

Read from the source code; the running server may differ. Descriptions are the server’s own words; hidden characters are shown as [U+…].

• audit_log

  Return recent outbound calls (method, host, path, status) for auditing.

• execute

  Execute a loaded operation. Credentials are injected from environment variables / keyring, the target host must be allowed by the security policy, and the response is size-capped. Args: operation_id: Id from `search_operations` (e.g. ``github.repos_get``). params: Object matching the operation's ``input_schema`` (path/query/header values plus an optional ``body``). extract: Optional list of dotted paths to keep from the response data, discarding everything else (saves context). ``*`` fans out over arrays: e.g. ``["items.*.name", "total_count"]``. fresh…

• execute_chained

  Execute a sequence of operations in one call, passing data between steps. Each step is an object ``{"operation_id", "params"?, "save_as"?, "extract"?, "fresh"?, "stop_on_error"?}``. A later step can reference an earlier result with ``${save_as.path.to.value}`` inside its params (e.g. ``${step1.data.id}``). A nested LIST of step objects forms a parallel group: those steps run concurrently (e.g. query three APIs at once) and later steps can reference all of their results. Use `extract` per step to keep responses small. Args: steps: Ordered list of step objects and/or parallel groups (list…

• execute_graph

  Execute a dependency graph of operations, auto-parallelizing independent ones. Prefer this over `execute_chained` when steps have a mix of dependencies: you do not order the steps or group parallel ones by hand. Instead each node declares an ``id``, and the engine infers the execution order from the ``${id.path}`` references in each node's params - running everything that is ready at the same time. Each node is ``{"id", "operation_id", "params"?, "extract"?, "fresh"?, "depends_on"?}``. A ``${other.data.x}`` reference in params makes this node wait for ``other``; ``depends_on: ["id"]`` adds a…

• get_operation

  Get the full input schema, response schema and auth requirements of one operation.

• list_apis

  List all currently loaded APIs with their protocol, base URL and size.

• load_api

  Load an API into the connector from a spec URL, file path, or raw content. The protocol (openapi/graphql/grpc/soap) is auto-detected unless given. After loading, use `search_operations` to discover operations. Args: spec: URL, local file path, or raw spec text. name: Optional short name used to namespace operations (``<name>.<op>``). base_url: Override the API base URL if the spec omits or misstates it. protocol: Force a protocol instead of auto-detecting.

• search_catalog

  Search the built-in catalog of free/public APIs ready to be loaded. Covers a curated list (GitHub, Stripe, OpenAI, Wikipedia, weather, ...) plus the APIs.guru directory of 2500+ public OpenAPI specs (Google, AWS, Microsoft, Twilio, ...). Each result contains a `spec` URL (and sometimes `base_url`/`protocol`) to pass straight to `load_api`. Args: query: Keywords describing the API you need (e.g. "weather forecast"). limit: Maximum number of results. include_directory: Also search the APIs.guru directory (fetched once and cached), not just the curated list.

• search_operations

  Fuzzy-search operations across loaded APIs. Returns operation ids you can pass to `get_operation` or `execute`. Args: query: Keywords to search for (matched against id, path, summary, tags). api: Restrict the search to a single loaded API by name. limit: Maximum number of results.

• unload_api

  Remove a previously loaded API and free its operations.

Compare and alternatives

Featured in

• Best MCP servers for GraphQL

Signals

Stars

4

Forks

0

Open issues

0

Weekly downloads

44

Last commit

Jul 24, 2026 (3 months ago)

Last release

Jul 23, 2026 (3 months ago)

Last package publish

Jul 23, 2026 (3 months ago)

From GitHub (fetched Oct 24, 2026) and the package registry.

Details

Registry name

io.github.TeodorMCP/universal-connector-mcp

Version

0.2.0

Auth

API key

Packages

pypi universal-connector-mcp

License

MIT

Repository

github.com/TeodorMCP/universal-connector-mcp

First seen

Oct 24, 2026

Last checked

Oct 24, 2026

Source

Official MCP Registrygithub-graphql