> For the complete documentation index, see [llms.txt](https://docs.hivel.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.hivel.ai/using-hivel/connecting-to-hivel-via-mcp.md).
# Connecting to Hivel via MCP
Hivel MCP lets you bring your engineering metrics directly into the AI assistant you already use for work — Claude, Cursor, or any other MCP-compatible client.
Instead of logging into Hivel to pull up a dashboard, you can ask your AI assistant questions about your Git and Jira metrics in plain language, right from the tool where you're already writing code or planning work.
Hivel MCP is currently available for read-only queries against your Git and Jira metrics. It does not modify data in Hivel or your connected tools.
***
### What Hivel MCP Helps With
#### **Team & People Performance**
**Team performance** *(team metrics)* How is a team performing? Any engineering measure — speed, quality, throughput, review culture — as a snapshot, a trend over time, or teams side by side.
* How is the Platform team's cycle time trending this quarter?
* Compare PR volume across our three backend teams last month.
* What's our overall review coverage this year?
**Developer performance** *(developer metrics)* How is one person doing? The same measures broken down per developer — great for 1:1s, spotting overload, or seeing individual contribution within a team.
* How many commits did Sri make last week?
* Show each developer's cycle time on the Payments team.
* Who reviewed the most pull requests this sprint?
**Repository health** *(repository metrics)* How is a codebase doing? Performance per repository — which repos are most active and how they compare.
* Which repositories had the most activity last month?
* Compare cycle time across our top five repos.
**Working cadence** *(active days)* How consistently are people working? The number of days developers were actually active in code — useful for cadence, availability, and workload.
* How many active days did the team average last month?
* Show active-day trends across the quarter.
#### **Code Activity**
**Commits** *(commits)* What was actually worked on? Individual commit records — what changed, and the mix of new work, rework, and maintenance.
* Show the commits on the checkout service this week.
* What did the team work on in the last three days?
**Pull requests** *(pull requests)* What's in flight, and how is it moving? Individual PRs to browse, plus where time goes across the PR lifecycle and which repos are busiest.
* List the open PRs for the Growth team.
* Where is time going in our PR process — coding, review, or merge?
#### **Delivery & Reliability**
**Release and Hotfix Health**
How healthy are your releases? Track hotfix rate and resolution time, release frequency, and code quality metrics like unreviewed PRs shipping to production. Understand the stability of your releases and how quickly your team responds to urgent fixes. Track release cadence and identify patterns in your deployment frequency. Measure code quality gatekeeping and identify review process gaps.
* What's our hotfix rate, and how long do hotfixes take to resolve?
* How often do we merge to release branches?
* What share of PRs got merged without a review last quarter?
#### **Project & Sprint Delivery (Jira)**
**Work items** *(work items)* How is work distributed? Counts and charts of Jira work by type, status, team, or sprint — plus how effort splits between features, bugs, and maintenance.
* How many stories versus bugs are in the current sprint?
* Show the split between new features and maintenance work.
**Individual tickets** *(issues)* What are the specific tickets? Browse Jira tickets with status, assignee, and type — and the commits or PRs linked to each.
* List the in-progress tickets for the mobile team.
* Which tickets were completed last sprint?
**Sprint health** *(sprint metrics)* How healthy is a sprint? Sprint summaries and hygiene scores — how well it's tracking and where data quality slips.
* How is the current sprint going?
* What's the hygiene score for last sprint?
**Planned vs. delivered** *(delivery)* Did we ship what we planned? Planning accuracy per sprint — say-do ratio, spillover, and delivery trends.
* How much of our planned work did we ship last sprint?
* Show planning versus delivery for Q2.
**Process hygiene** *(hygiene)* Is our process clean? Data-quality checks — PRs and commits not linked to tickets, and tickets missing estimates, owners, or gone stale.
* Which PRs aren't linked to a Jira ticket?
* Show tickets that are missing estimates.
**Behind the Scenes**
**Automatic lookups** *(org & Jira context)* Two helper capabilities the assistant uses on its own to find your teams, people, repositories, boards, and sprints — so "Sri's cycle time" becomes the right person on the right team before it answers. You never need to ask these directly.
***
### How to Connect
MCP connection details can be found under [Profile](https://app.hivel.ai/settings/user-profile) → MCP Connection. Follow the steps, and you're ready to query Hivel from your own AI instance.
You'll need your MCP connection details. These include a server URL and a set of headers unique to your account, including your access token.
#### Claude Connector Settings
You can also connect using Claude's connector settings directly. In Claude, go to Settings → Connectors and add Hivel as a custom connector using the server URL from your Hivel MCP Connection tab. Use the code given below
```
https://mcp.hivel.ai/hivelmcp/mcp
```
Log in to Hivel, and use the Claude connector to connect Hivel to Claude.
Keep in mind that if you log out of Hivel, you'll need to repeat all of the steps above to reconnect.
***
### A Note on Sessions
Hivel MCP access is tied to your active AI client session, not stored permanently on Hivel's side.
* As long as you stay logged into your AI instance (Claude, Cursor, etc.), the connection stays live and you can keep asking questions.
* If you log out of your AI client, access is lost, and you'll need to reconnect MCP by repeating the setup steps above.
This is expected behavior — it's the same session-based model most MCP integrations follow, and it keeps your credentials from persisting somewhere outside your control.
***
### Example: Here is how data will appear on your screen
* Prompt: Can you give me the commit count for all teams in Q2 and compare it with Q1
* Prompt: Please provide details of Work Items closed by X Engineer in the current sprint
***
### When to Use MCP vs. Cockpit vs. Lucy
**Use MCP when you want to:**
* Ask about metrics without leaving your coding or planning tool
* Get a quick answer mid-conversation with your AI assistant
* Combine Hivel data with other context already in that conversation (e.g., "given this PR, what's our team's usual review time?")
**Use Cockpit dashboards when you need:**
* Deep drilldowns or detailed PR-level analysis
* Custom filters or precise metric investigation
* Comprehensive sprint planning or historical trend analysis
**Use Lucy when you want to:**
* Ask Hivel-native questions directly inside the Hivel platform without any setup
***
For more information or assistance, please contact our support team at .