Cline MCP server: it files bug reports

An official server that does one thing nobody expects, why that makes Cline a client rather than a data source, and what to connect instead.

Last verified 27 August 2026 · from The cline-community repository README and the GitHub API

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 Cline MCP server?

Cline is an AI coding agent that runs inside an editor. Like Cursor and Windsurf, its main relationship to the Model Context Protocol is as a client: you configure servers and Cline calls their tools.

The one server it publishes points inward at its own bug tracker. It exists so that a user hitting a problem can have Cline file a well-formed GitHub issue without leaving the editor.

That explains the shape of the tools. The value is the automatic context gathering. A bug report that carries the operating system, the Cline version, the API provider and the model is more actionable than one without them.

Cline works in code, tests and review, and its server reaches only its own issue tracker.

Cline is a consumer, not a source. The one server it ships points at GitHub, not at Cline. An agent that wants to know what a Cline task did must read the workspace or the git history instead.

What can an AI agent do with Cline?

What connecting costs before the first question

A context window is the amount of text a model can hold at once. Two tools, and both of them file GitHub issues about Cline rather than reaching Cline data. The definitions cost 472 to 628 tokens on Claude, which is among the cheapest connections measured anywhere in this guide.

That number is small because the surface is small, not because the server is efficient. What it buys is issue filing, and the page above explains why that is the whole of it.

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 from the 2 tool entries in the vendor's own documentation. That documentation is the only public inventory.

The server could not be run for a live tools/list. No schema was captured from the server, so these figures cover each tool's name and description only. They are a floor, not the figure.

The Claude columns are measured through the API's count_tokens endpoint with an empty schema per tool. So they cover the same content as the GPT columns. The range there is two tokenizer generations. Haiku 4.5 and Opus 4.6 are at the low end. Opus 5 and Sonnet 5 are at the high end. The gap between them is about a third.

Cline MCP server limits: where answers come back incomplete

One limit below, and it is about what the server does not expose.

How to configure Cline MCP for agents

1. 1

   Install the GitHub CLI first

   The server shells out to gh and does nothing without it authenticated.

2. 2

   Build it yourself

   There is no npm package. The documented route is git clone, then a build, then pointing the config at build/index.js.

3. 3

   Auto-approve the preview, never the submit

   The README puts preview_cline_issue in autoApprove and leaves report_cline_issue out, which is the right split: one is local formatting, the other posts in public.

4. 4

   Know what it attaches

   Every issue carries your OS and release, the Cline version, the IDE, and the API provider and model read from the task metadata file.

5. 5

   Expect a 10-second timeout

   That is the value in the README's own configuration block.

Paste this into your agent’s instructions

The two tools differ by one word and by whether the result is public. The block makes the agent say which one it is about to use.

When you use the Cline community MCP tools, state:
- Which tool you are calling. preview_cline_issue formats
  an issue locally. report_cline_issue posts it publicly to
  the cline/cline GitHub repository.
- That the issue body will carry the operating system, the
  Cline version, the IDE, and the API provider and model
  read from the task metadata.
- Never call report_cline_issue without being asked to file
  an issue.
Never state that data does not exist. This server reads no
Cline data: it only files issues.
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?

No, for most teams. A repository connector files issues with maintenance behind it, and GitHub's own server exposes issue creation among 85 tools, though without Cline's automatic version and provider capture.

For what a Cline session actually did, read the result rather than the agent. A repository connector shows the commits, the diff and the branch, which is the durable record of the work.

For token spend across agents, Cline's own figures are not exposed, so a gateway or an observability product is the route. Helicone, Langfuse and Traceloop all reach model call records with cost attached.

For connecting servers to Cline itself, that is the normal direction and needs nothing from this page: Cline reads a standard client configuration.

Use a repository connector for issues and for what changed. Keep cline-community only if the automatic context capture is worth a dormant dependency.

Cline MCP server setup

There is no package to install. The README's route is to clone the repository, build it, and point the configuration at the built entry point.

The credential is a GitHub token in GH_TOKEN. The README obtains it by running gh auth login, choosing the browser flow, and then reading it back with gh auth token.

This server is configured inside Cline itself rather than in a general MCP client, so the file path depends on which editor hosts Cline.

gh repo clone cline/cline-community
cd cline-community
npm install && npm run build

gh auth login      # choose: Login with a web browser
gh auth token      # copy this into GH_TOKEN

{"mcpServers":{"cline-community":{
  "autoApprove":["preview_cline_issue"],
  "timeout":10,
  "command":"node",
  "args":["/path/to/cline-community/build/index.js"],
  "transportType":"stdio",
  "env":{"GH_TOKEN":"<your-token>"}}}}

{"mcpServers":{"cline-community":{
  "autoApprove":["preview_cline_issue"],
  "timeout":10,
  "command":"node",
  "args":["C:\path\to\cline-community\build\index.js"],
  "env":{
    "APPDATA":"C:\Users\<username>\AppData\Roaming",
    "GH_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.

claude mcp add-json "cline-community" '{"command":"node",
  "args":["/path/to/cline-community/build/index.js"],
  "env":{"GH_TOKEN":"<your-token>"}}'

{"mcpServers":{"cline-community":{
  "command":"node",
  "args":["/path/to/cline-community/build/index.js"],
  "env":{"GH_TOKEN":"<your-token>"}}}}

Read-only access, permissions and security

• One tool posts in public

  report_cline_issue creates an issue on the cline/cline repository through your authenticated GitHub CLI, under your own account.

• The issue body carries an environment fingerprint

  The README lists what is attached automatically: operating system and release, Cline version, IDE, and the API provider and model taken from the task metadata file.

• The API provider and model come from task metadata

  That is a record of which vendor and model you were using, published into a public issue.

• The token is a GitHub token in plain configuration

  GH_TOKEN sits in cline_mcp_settings.json, so it carries whatever scope your gh login has.

• The repository carries no license

  GitHub reports none for cline/cline-community, which is a question for anyone whose policy requires one.

Troubleshooting

`gh: command not found`

The server shells out to the GitHub CLI. The README lists it as a requirement, installed and authenticated, alongside access to the task metadata directories.

The issue has no API provider or model

Those fields are read from Cline's task metadata file. Outside a Cline task, or where the metadata directory is unreadable, there is nothing to extract.

Metadata is missing on Windows

Set APPDATA in the configuration, and put it in the same env object as GH_TOKEN. The README's Windows example declares env twice, so one of the two values is discarded before the server ever starts.

The server times out

The README's configuration sets timeout to 10. Filing an issue goes over the network through gh, so a slow connection can exceed that.

You cannot find it on npm

It is not published. The documented install is a clone and a local build, and the configuration points at build/index.js by absolute path.

Cline MCP server: Reference

What engineers report

cline/cline-community is first-party, created on 11 April 2025, with 23 stars. The useful facts about it are its purpose and its dates.

The cline GitHub organization and the cline-community README were read on 25 August 2026. The README describes the server as simplifying issue reporting from Cline to GitHub, and lists the system information it gathers. No tool exposing Cline's own task or cost data was found.

Should you connect an agent to Cline?

Yes, though it is no longer actively maintained. Best for filing well-formed Cline bug reports from the editor, with report_cline_issue kept out of autoApprove.

Strongest fit

Cline users who want well-formed bug reports filed from the editor with version and provider attached.

Main advantage

Automatic context capture: operating system, Cline version, API provider and model, gathered without the user assembling them.

Main weakness

It exposes none of Cline's own data, so it is not a connector to Cline in the sense most searches mean.

Operational risk

A dormant, unlicensed dependency that writes to your GitHub repository.

Frequently asked questions