Autter

Your agent can finally talk to its reviewer.

The Autter MCP server connects coding agents to everything Autter knows about your code: review findings, team conventions, blast radius, and dependency risk. The agent fetches its review, fixes the findings, and asks for another pass without leaving the editor.

Claude Code · Cursor · OpenCode · Windsurf · anything that speaks MCP

claude "address the autter review on pr 142"

get_pr_reviewcheckout-service #142

3 findings · 1 blocking · changes requested

fix_prompt · refund.ts:88 · missing idempotency key

✎ patched payments/refund.ts · retry now idempotent

reply_to_review_comment"fixed in a41c9e2"

resolve_review_thread✓ resolved

trigger_reviewqueued · revision 3

review round-trip complete · 0 dashboard visits

Review stops being a handoff.

The old loop had a human courier: read the finding on GitHub, explain it to the agent, paste the fix back, click resolve. Over MCP the agent runs the whole circuit itself.

  1. 01

    Pull the review

    get_pr_review returns the verdict and up to 50 findings, each pinned to file and line with a ready-to-run fix_prompt.

  2. 02

    Fix in place

    The agent applies each fix_prompt in the working tree, with the conventions from get_learnings already in context.

  3. 03

    Answer the thread

    reply_to_review_comment posts the fix commit; resolve_review_thread closes the thread out on GitHub.

  4. 04

    Push

    Commits go up exactly the way they always do. Nothing wraps Git.

  5. 05

    Ask for another pass

    trigger_review re-runs the latest revision. Green means merge.

Sixteen tools, five jobs.

Not an API dump behind an MCP wrapper. Each tool is built around something an agent genuinely needs mid-change, from identity to impact analysis.

Every call states the reason it is calling, and every call lands in your organization's audit log.

Read the launch announcement

Identity and repositories

The first calls of any session.

whoami
The authenticated user, org, token, and connected repositories. Call it first.
list_prs
Tracked pull requests for a repo, newest first, with per-PR review status.

The review loop

Everything needed to take a review from findings to green.

get_pr_review
The verdict, the chapters, and up to 50 findings, each pinned to file and line with a ready-to-run fix_prompt.
reply_to_review_comment
Answer a review thread as the agent, attributed to your token.
resolve_review_thread
Close a thread once its finding is fixed. GitHub stays the source of truth.
trigger_review
Re-run the review on the pull request's latest revision.

Team knowledge

The conventions your reviews have been teaching Autter all along.

get_learnings
Team conventions learned from review feedback, dismissed findings, and repo rules.
add_learning
Teach Autter a new repo convention. Future reviews enforce it.

Codebase intelligence

Answers before the first edit, from the graph that already mapped the repo.

query_codebase
Natural-language questions answered from generated docs and the code graph, with citations.
get_symbol_info
Definition, callers, blast radius, and change risk for any symbol.
get_impact_analysis
What a change to a file can reach: dependents, churn history, test coverage.
get_repo_health
Architecture concerns, hotspots, and risk insights at repository level.

Security and scans

The checks that usually arrive too late, moved before the decision.

check_dependency
Known CVEs, fixed versions, and typosquat flags before the package is installed.
trigger_scan
Start a full codebase scan across Autter's scanning agents.
get_scan_status
Poll a running scan with per-agent progress.
get_scan_findings
Unified findings from the latest scan, filterable by severity and agent.

Connected in one command.

Create a token, point your client at the endpoint, and the next session starts with the reviewer in the room.

  1. 01

    Create a token

    In the Autter dashboard, open Organization, then Access Tokens. Scope it to mcp:read for a read-only agent.

  2. 02

    Add the server

    One command for Claude Code, one JSON block for Cursor and everything else.

  3. 03

    Put it to work

    Ask your agent what Autter found on the open pull request. It knows where to look now.

Claude Code

claude mcp add --transport http autter https://api.autter.dev/mcp \
  --header "Authorization: Bearer autter_pat_..."

Cursor and other MCP clients

{
  "mcpServers": {
    "autter": {
      "type": "http",
      "url": "https://api.autter.dev/mcp",
      "headers": { "Authorization": "Bearer autter_pat_..." }
    }
  }
}

Streamable HTTP at https://api.autter.dev/mcp, authenticated with an Autter personal access token.

Built with the doors locked.

A machine-facing surface earns trust the same way a reviewer does: by being predictable about what it will and will not do.

Scoped, revocable tokens
Personal access tokens carry mcp:read, mcp:write, and mcp:llm scopes, expire on schedule, and revoke from the dashboard.
Every call audited
Actor, token, duration, result, and error code, recorded in the organization audit log.
Rate limited per token
Runaway agent loops get slowed down instead of taking the org down.
No side doors
No arbitrary SQL, no generic GitHub writes. Every operation is scoped to Autter's review, scan, and code-intelligence workflows.

Close the loop

Let your agent answer its own review.

Create an access token, add one command to your agent, and the next review round-trips without a human courier.

Connect your agentStart Now