Skip to content
New: the GitDailies MCP server

Engineering Metricsfor Your AI Agents

Connect Claude, ChatGPT, Cursor, or GitHub Copilot to GitDailies. Ask about PRs, reviews, and DORA metrics - and get answers from real data.

Connect your agent Book a demo
An AI agent calling GitDailies MCP tools to explain a spike in weekly PR pickup time

MCP server

Ask your agent.Get answers from real data.

The GitDailies MCP server gives AI agents the same PR metrics, DORA metrics, review queues, and alerts you see in GitDailies - with drill-downs to the exact PRs behind every number.

Your review queue

What needs me today?

An AI agent listing three pull requests waiting for review, oldest first

Your agent sees your open PRs and outstanding review requests - including those requested from your teams - so it can tell you what to pick up first.

DORA metrics

DORA in plain language.

An AI agent summarizing deployment frequency, lead time, change failure rate, and time to restore for the quarter

Ask how delivery is trending by week, month, or quarter, per repo or across the org. Your agent gets the four DORA metrics straight from GitDailies.

Investigations

Explain Every Spike.

When a metric moves, your agent can drill into the exact PRs, reviews, deploys, or incidents behind any week or month.

Filter and split by repo, team, user, or label, exclude bots and weekends, and get the same numbers as your GitDailies dashboards.

An AI agent listing the pull requests behind a spike in DORA lead time for change, longest first

Standups and updates

Status updates in seconds.

An AI agent drafting a standup update from merged PRs, submitted reviews, and open PRs

Your agent pulls what you merged, what you reviewed, and what’s still open, then drafts your standup, weekly update, or retro notes.

Alerts

Catch up on what fired.

An AI agent listing two GitDailies alerts that triggered overnight

Your personal GitDailies alerts - huge PRs, critical files changed, PRs stuck in review - are available to your agent, with the events that triggered them.

Setup

Works With Your Agents.

Claude, Claude Code, ChatGPT, Cursor, VS Code with GitHub Copilot - or any MCP client, including agents you build yourself.

Add one URL and sign in with OAuth. Hosted for you: no API keys, nothing to install, nothing to run.

Read the setup guide
Adding the GitDailies MCP server with the Claude Code CLI and in a Cursor mcp.json config file

Why GitDailies

GitHub has an MCP server.So why does your agent need GitDailies?

GitHub's MCP server is built for working with code: reading files, opening issues, managing PRs. But engineering metrics aren't sitting in GitHub waiting to be read. They have to be collected, joined, and computed - and that's exactly what GitDailies does.

Your agent + raw GitHub data

  • Pages through thousands of PRs, reviews, and timeline events, one API call at a time
  • Fills its context window - and your token bill - with raw JSON
  • Runs into API rate limits long before it reaches last quarter
  • Does the date math and averaging itself, and can get it wrong
  • Can give a different number every time you ask

Your agent + GitDailies

  • Asks for "weekly pickup time, by team, for the last 6 months" in a single call
  • Gets compact, precomputed metrics that are ready to reason about
  • Up to 18 months of history, across every repo you follow
  • Every number computed by tested, deterministic code
  • The same answer as your GitDailies dashboards and reports, every time

  • 01

    Don't sip from the firehose.

    GitDailies captures every PR, review, and deploy event as it happens, and structures it into metrics and observations designed for agents to query - not raw API payloads.

  • 02

    Computed, not guessed.

    Pickup time, lead time for change, change failure rate: each is derived from several events across PRs, reviews, commits, deploys, and incidents. GitDailies computes them accurately and reliably, by the same definitions every time.

  • 03

    No hallucinated stats.

    LLMs are brilliant at spotting patterns, and unreliable at arithmetic over thousands of rows. GitDailies does the heavy lifting of gathering and computing metrics; your agent does what it’s best at - finding the patterns, explaining causes, and adding color and insight.

  • 04

    Slice it any way you like.

    Split and filter by team, repo, user, label, and branch. Exclude bots, drafts, small PRs, and weekends in your time zone. Bucket by day, week, month, or quarter. GitHub’s API doesn’t aggregate any of this - GitDailies exposes it all in a form your agent can use effortlessly.

  • 05

    Your alerts, included.

    GitDailies watches your GitHub activity with custom alert rules - huge PRs, changes to critical files, stalled reviews. Your agent can see which alerts fired, and the events that triggered them.

  • 06

    Every number is traceable.

    Drill-down tools return the exact PRs, reviews, and deploys behind any data point. When your agent tells you lead time spiked, it can show you why - and you can check its work.

Better together. Keep GitHub’s MCP server for working with code, and add GitDailies for answers about how your team delivers.

Connect your agent in a minute

The GitDailies MCP server is hosted for you. There’s nothing to install or run. Add this URL to any MCP client that supports remote servers, and sign in with GitDailies when prompted:

https://mcp.gitdailies.com/mcp

To get started first install the GitDailies App for GitHub in your GitHub organization, then follow the step-by-step setup guide for Claude, ChatGPT, Cursor, VS Code, and other clients. See the tools reference for everything your agent can do, and example prompts to try first.

MCP access is included in the Max plan. Contact support@gitdailies.com for a free trial.

Privacy and security

The MCP server is read-only: your agent can query metrics, but can’t change anything in GitHub or GitDailies. Each connection signs in with OAuth and is bound to a single GitHub organization, and your agent only sees the repos that you follow in GitDailies. Like the rest of GitDailies, the MCP server works from GitHub metadata only - it never reads your source code.

Please see our privacy policy and security page for full details.

Questions

MCP server FAQ

The Model Context Protocol is an open standard for connecting AI assistants and agents to data sources and tools. Most AI assistants and coding agents support it, including Claude, ChatGPT, Cursor, and GitHub Copilot.

GitHub’s MCP server gives your agent raw data: individual PRs, reviews, and events, fetched one API call at a time. Turning that into reliable metrics means paging through thousands of records, joining them, and doing the math - which burns tokens, hits rate limits, and invites hallucinated numbers. GitDailies has already collected and computed it all, with powerful filtering and aggregation, so your agent gets accurate metrics in one call and can focus on insight. It also gives your agent things GitHub doesn’t have at all, like DORA metrics and your triggered alerts. See the full comparison. The two work well together: use GitHub’s server for working with code, and GitDailies for metrics.

MCP access is included in the Max plan. Contact support@gitdailies.com for a free trial.

PR, review, and DORA metrics for the repos you follow in GitDailies, the PRs and reviews behind them, and your own open PRs, review requests, recent work, and personal alerts. See the tools reference for the full list.

No. The MCP server is read-only. It can’t comment on, approve, or merge PRs, and it can’t change your GitDailies settings.

No. Like the rest of GitDailies, it works from GitHub metadata only - PR titles, reviewers, timestamps, and status. GitDailies never clones your repos or reads your source code.

Yes. Each connection is bound to one organization, so add one connection per organization. The setup guide explains how.

Give your agents context.Ship with confidence.

Install GitDailies in 90 seconds, then connect your AI agent with one URL.

Try for Free Connect your agent