Codacy MCP server: 24 tools for code quality and security

Codacy keeps code-quality findings and security findings apart for an agent. The tool list has a measured cost before the first query.

Last verified 1 September 2026 · from Codacy's repository, its documentation and the server's own tools/list response

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 Codacy MCP server?

It is Codacy's server for the Model Context Protocol. Codacy analyzes repositories for quality, coverage and security, so the agent can query those findings.

The surface splits three ways. Repository tools read issues, files, coverage and duplication. Pull-request tools read the diff, its issues and its coverage. SRM tools, for Software Risk Management, read security findings, which Codacy keeps separate from quality issues.

Two further tools run analysis on your machine rather than reading stored results, by installing and invoking the Codacy CLI.

Codacy covers code, tests and review broadly, reaches part of the way into release through pull-request gating, and reaches nothing in production observability, agent observability or cost.

In practice, Codacy answers what is wrong with the code and what the tests cover, and nothing about what the code does once it runs. An agent connected here can find the security findings on a branch and the files with no coverage. It cannot tell you whether any of it matters in production, so this connector pairs with a telemetry one.

What can an AI agent do with Codacy?

What connecting costs before the first question

A context window is the amount of text a model can hold at once, and connecting uses some of it before the agent asks anything. Every tool the server registers sends its name, description and input schema to the model when the session starts. The tool list is charged whether or not a single tool is called.

Codacy's descriptions average 591 characters, longer than most, because several carry a common mistakes section: guidance written for the agent about how the tool is usually misused. That is unusual, and it is charged at connection like everything else.

Encodings: GPT-5.6, GPT-5.5 o200k_base; GPT-4, GPT-3.5 cl100k_base; Claude Haiku 4.5 to Opus 5 count_tokens. Counted 24 August 2026 from tools/list on the published @codacy/codacy-mcp package. It was run locally with a placeholder token.

Registration is static, so no Codacy account was involved. Figures cover name, description and input schema.

The two models in column one share the o200k_base encoding, so they get one count. The spread is formatting, about 21%. Reasoning tokens are output, spent per request. The tool list is input, charged once at connection.

The Claude figures are measured rather than encoded. Anthropic publishes no local tokenizer. So the definitions were sent to the API's count_tokens endpoint, and a no-tools baseline was subtracted. The range covers two tokenizer generations rather than two serializations. Haiku 4.5 and Opus 4.6 sit at the low end. Opus 5 and Sonnet 5 sit at the high end. The gap between them is about a third.

Codacy MCP server limits: where answers come back incomplete

Codacy's first limit is not about data at all. Two of its tools download and run a binary on your machine.

How to configure Codacy MCP for agents

1. 1

   Decide whether the agent may run the CLI

   codacy_cli_install and codacy_cli_analyze fetch and execute the Codacy CLI locally. That is how unpushed code gets analyzed, and it is also the only code execution here.

2. 2

   Pin CODACY_CLI_VERSION

   Without it the fetched version is whatever is current, so the same analysis can give different results on different days.

3. 3

   Tell the agent which tool answers security questions

   Codacy separates quality issues from SRM security findings, and the wrong tool returns a confident, empty answer.

4. 4

   Budget the context

   24 tools cost about 12,072 tokens, roughly 6% of a 200,000-token window.

Paste this into your agent’s instructions

Two of Codacy's failure modes are indistinguishable from good news: the wrong tool for security, and missing coverage uploads. Both return empty. The block makes the agent name what it queried.

When you answer using Codacy MCP tools, state:
- Which tool you used. Quality issues and security findings
  are separate surfaces, and only the SRM tools answer
  security questions.
- The repository and branch, and whether coverage reports
  exist for it. Missing uploads look like missing tests.
- Whether you ran the local CLI, which executes on this
  machine rather than reading stored analysis.
Never state that data does not exist. State that none was
returned under the constraints above.
Write your answer in ASD-STE100 Simplified Technical
English. Use short sentences with one idea in each.

Do you need the MCP server at all?

Partly. Codacy has a documented REST API, and a coding agent can call it with the same account token.

The connector adds two things the API does not. First, the tool descriptions carry usage guidance, including which tool answers a security question rather than a quality one. That is the mistake an agent most often makes here. Second, the CLI tools analyze code that has not been pushed, which no hosted API can do.

For reading stored results on a schedule, the API is lighter than 12,072 tokens of definitions.

Connect the server for interactive review, especially on unpushed code. Use the API for scripted reporting.

Codacy MCP server setup

The package runs with npx and needs a Codacy account token in CODACY_ACCOUNT_TOKEN.

CODACY_CLI_VERSION pins the CLI that local analysis downloads. Leaving it unset means whatever is current at the time.

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

{"mcpServers":{"codacy":{"command":"npx",
  "args":["-y","@codacy/codacy-mcp"],
  "env":{"CODACY_ACCOUNT_TOKEN":"<your-token>",
         "CODACY_CLI_VERSION":"<pinned-version>"}}}}

[mcp_servers.codacy]
command = "npx"
args = ["-y", "@codacy/codacy-mcp"]
env = { CODACY_ACCOUNT_TOKEN = "<your-token>" }

Every other client

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

{"mcpServers":{"codacy":{"command":"npx",
  "args":["-y","@codacy/codacy-mcp"],
  "env":{"CODACY_ACCOUNT_TOKEN":"<your-token>","CODACY_CLI_VERSION":"<pinned-version>"}}}}

{"mcpServers":{"codacy":{"command":"npx",
  "args":["-y","@codacy/codacy-mcp"],
  "env":{"CODACY_ACCOUNT_TOKEN":"<your-token>","CODACY_CLI_VERSION":"<pinned-version>"}}}}

{"servers":{"codacy":{"command":"npx",
  "args":["-y","@codacy/codacy-mcp"],
  "env":{"CODACY_ACCOUNT_TOKEN":"<your-token>","CODACY_CLI_VERSION":"<pinned-version>"}}}}

{"mcpServers":{"codacy":{"command":"npx",
  "args":["-y","@codacy/codacy-mcp"],
  "env":{"CODACY_ACCOUNT_TOKEN":"<your-token>","CODACY_CLI_VERSION":"<pinned-version>"}}}}

Read-only access, permissions and security

• Two tools execute code on your machine

  codacy_cli_install downloads the Codacy CLI and codacy_cli_analyze runs it. Every other tool reads a hosted API.

• Pin the CLI version

  An unpinned fetch means the binary can change between runs, which is both a reproducibility and a supply-chain consideration.

• The token carries your Codacy access

  An account token reaches the organizations and repositories that user can see.

• One tool changes your account

  codacy_setup_repository adds or follows a repository in Codacy, which is the only hosted-side write here.

Troubleshooting

No security issues found on a repository that has them

The quality tool was probably used. Security findings live behind codacy_search_repository_srm_items and codacy_search_organization_srm_items, which Codacy's own descriptions name as the primary security tools.

Coverage is empty everywhere

Coverage tools read reports uploaded to Codacy. A repository whose pipeline never uploads returns empty, which is not the same as untested code.

The server hangs on first use

A local-analysis tool triggers a CLI download on first call, which takes longer than an ordinary request. Subsequent calls reuse it.

Local analysis gives different results than last week

The CLI version is not pinned. Set CODACY_CLI_VERSION so the same analysis runs each time.

Codacy MCP server: Reference

What engineers report

codacy/codacy-mcp-server has 62 stars and was last pushed on 30 July 2026, with the package at 0.6.24. Public discussion is limited to a handful of issues, two of which are quoted below.

“When a tool is configured to use the repository configuration file, ... the MCP's codacy_cli_analyze still regenerates the Codacy pattern-catalog config and analyzes against it, ignoring the repo's config file. This produces false positives.”

“I see there're .cursor/rules/codacy.mdc file and .codacy/ folder. But the agent report error. zsh: command not found: codacy_cli_analyze”

Repository facts from the GitHub API and package facts from npm, both retrieved 24 August 2026. Tool counts and description lengths come from the server's own tools/list response.

Should you connect an agent to Codacy?

Yes. Best for codacy teams pinning the CLI version, or leaving the CLI tools off the allowlist entirely.

Strongest fit

Teams on Codacy who want an agent reading quality, coverage and security findings while reviewing a pull request.

Main advantage

A clear split between quality and security, with tool descriptions that steer an agent to the right one.

Main weakness

No telemetry, so findings arrive without any evidence of whether they matter in production.

Operational risk

Two tools download and run a CLI on your machine, which is a different kind of trust from a read-only API.

Frequently asked questions