GitHub Issues MCP server

How an agent would call it

Two steps: search returns refs, then data_fetch opens one of them. The arguments below are the arguments that were sent.

{
  "tool": "search",
  "arguments": {
    "query": "repo:modelcontextprotocol/servers is:issue authentication",
    "sources": [
      "github-issues"
    ],
    "limit": 2,
    "runId": "<from run_begin>"
  }
}

{
  "tool": "data_fetch",
  "arguments": {
    "sourceId": "github-issues",
    "ref": {
      "id": "modelcontextprotocol/servers#3047"
    },
    "runId": "<from run_begin>"
  }
}

This runs against an account — send Authorization: Bearer <token> with the call. Get one at app.bedrocktap.com/dashboard/tokens; the free tier needs no card, and it reaches github-issues like every other source.

What it costs. 0.45 credits per issue returned, and 0.45 per fetch. The search above returned 2 issues and cost 0.9 credits; the fetch cost 0.45.

What came back

{
  "ref": {
    "id": "modelcontextprotocol/servers#3047",
    "url": "https://github.com/modelcontextprotocol/servers/issues/3047",
    "title": "Register MCP reference and docs servers in the official server registry"
  },
  "text": "## Overview\n\nThis document tracks the registration of all official MCP reference servers and the documentation server in the [MCP Server Registry](https://registry.modelcontextprotocol.io/).\n\n**Goal**: Register all 8 servers to promote discoverability through the official registry as we deprecate README-based server listings.…",
  "fetchedAt": "2026-09-12T14:57:01.847Z"
}

Each ref carried repo, number, state, comments, author, createdAt and updatedAt in its meta. The fetched payload’s text was the issue body in full — 12,012 characters — and its data carried the whole issue record, including labels, assignees, reactions, state_reason, closed_at and sub_issues_summary.

Observed 2026-09-12. One call against a live source — ask again tomorrow and the rows move.

What a row contains

One row is one issue: an owner/repo#number id that a fetch resolves on its own, the html link, and the title. meta carries the repository, the issue number, open or closed state, the comment count, the author’s login and both timestamps. Pull requests are filtered out before the rows are built, because a proposal is not a report. A fetch returns the body as text and the full issue record as data.

A question it answers. What do this project’s users report broken, and how long do those threads stay open?

Routing guidance

What an agent reads to decide whether github-issues is the right source, and what to reach for when it is not.

Answers

What is broken or missing in a named project. The sharpest source here for bug reports and feature requests. Fetching a result returns the issue body plus its comments.

Query shape

GitHub search syntax works verbatim — `repo:owner/name`, `label:bug`, `state:open`, `reactions:>10`, `created:>2026-01-01`. A configured default repo list is appended when your query has no repo qualifier.

Price

Priced in credits, per row, per query or per call depending on the source. Every call is quoted before it runs — `cost_estimate` returns the exact figure and spends nothing. See /pricing.

Latency

Fast (~1-3s).

Prefer instead

`github-discussions` for open-ended Q&A and ideas. `stackexchange` when the target is a technology rather than a named project.

Cannot answer

Private repos, code search, or anything outside GitHub.

Caveat

A token is optional but lifts the rate limit substantially — without one, a wide sweep will start failing.

Related sources