Sourcery MCP server: there isn't one

What was searched, what the open feature request says, and which code-review connectors reach an agent.

Last verified 1 September 2026 · from Sourcery's documentation, its GitHub organization and the official MCP registry

This page is one of 90 in a directory of monitoring and developer tools. Each page checks what an AI coding agent can get out of the tool through its MCP server, the connector that lets the agent query the tool directly.

What is the Sourcery MCP server, and is it usable?

Sourcery is an automated code-review tool that comments on pull requests and suggests refactors. That makes it a natural candidate for an MCP server: an agent that could read Sourcery's findings would not have to re-derive them.

No such server exists. The search below covers the vendor's own documentation, its GitHub organization and the protocol's official registry, and all three are empty.

The open feature request matters because it separates two situations. A vendor that has never been asked is different from one that has been asked and has not acted, and Sourcery is the second.

Every one of the 90 vendors in this directory gets the same four checks. Sourcery passes 0 of them.

Sourcery works in code review, and publishes no connector, so its findings stay in the pull request it comments on.

Sourcery and your coding agent do the same job without talking to each other. Sourcery reviews the diff and comments on the pull request; your agent reviews the diff in the editor. Neither can read the other's conclusions, so the work happens twice.

What can an AI agent do with Sourcery?

An agent can do nothing with Sourcery directly, because no tool exists to call. One thing is worth watching, because search results make this look less settled than it is.

Do you need the MCP server at all?

No. Sourcery reviews pull requests on its own. For automated review findings an agent can read, Codacy exposes 24 tools over quality, coverage and security, and SonarQube documents 48 across analysis and issues.

For pattern-based review without a platform, Semgrep runs locally through its own CLI and costs among the least context measured in this guide.

Sourcery itself remains useful on pull requests. It simply is not something your agent can query.

Keep Sourcery on pull requests. Connect Codacy, SonarQube or Semgrep when the agent needs review findings.

Sourcery MCP server setup

Nothing here installs a Sourcery server, because there is none and the feature request has sat without activity since January 2026. This is the working route.

Keep Sourcery where it works: on pull requests, reviewing code. What it cannot do is answer an agent that asks what it found.

Three code-quality vendors in this directory publish servers that answer that question, and each takes a different shape. The commands below add them to Claude Code; Cursor takes the same values in ~/.cursor/mcp.json.

claude mcp add-json "codacy" '{"command":"npx",
  "args":["-y","@codacy/codacy-mcp"],
  "env":{"CODACY_ACCOUNT_TOKEN":"<your-token>"}}'

claude mcp add sonarqube \
  --env SONARQUBE_TOKEN=$SONAR_TOKEN \
  --env SONARQUBE_ORG=$SONAR_ORG \
  -- docker run --init --pull=always -i --rm \
     -e SONARQUBE_TOKEN -e SONARQUBE_ORG \
     sonarsource/sonarqube-mcp

claude mcp add-json "semgrep" '{"command":"semgrep",
  "args":["mcp"]}'

Every other client

Each block below is the configuration for one client, with the file path and the key that client expects.

Read-only access, permissions and security

• Sourcery requires no credential here

  The absence of a server means no Sourcery token is requested.

• Review findings describe your weaknesses

  A code-quality connector returns exactly where the code is fragile or insecure, which is worth scoping deliberately.

• Semgrep needs no account for local rules

  Running the CLI's server without a login is the narrowest of the three, since nothing leaves the machine to a vendor.

• SonarQube's container takes the token by environment

  Passing -e rather than embedding the value keeps the credential out of the image and out of shell history.

• Codacy's token carries repository scope

  What the agent can read is decided in Codacy, so scope the token before connecting rather than after.

Troubleshooting

You cannot find a Sourcery MCP server

There is not one. Sourcery's documentation index has no MCP entry, the sourcery-ai organization has no MCP-named repository, and the official registry returns nothing.

You found a feature request and want to know if it is progressing

It remains open, with no recorded activity since January 2026. Treat it as an unanswered request rather than as work in progress.

`pip install` or `npm install` finds nothing

Nothing was published. There is no package on npm or PyPI for a Sourcery MCP server.

You want a code-review tool an agent can query

Codacy, SonarQube and Semgrep all publish servers that answer that class of question. Each has its own page in this directory.

Sourcery MCP server: Reference

What engineers report

The searches are the evidence here, so they are recorded rather than summarized.

The sourcery-ai GitHub organization was searched for MCP-named repositories and the official MCP registry was searched for Sourcery, both on 1 September 2026. Both returned nothing.

Should you connect an agent to Sourcery?

No. There is nothing to connect. Best for pairing Sourcery's pull-request review with a code-quality connector the agent can query.

Strongest fit

Teams who want automated review comments on pull requests and do not need agent access.

Main advantage

None for an agent. Sourcery's findings stay in the pull request it comments on.

Main weakness

An agent cannot read Sourcery's findings, so review work is duplicated in the editor.

Operational risk

Waiting on the open feature request. It has had no recorded activity since January 2026.

Frequently asked questions