Keypup unifies the work in your Git, project and AI tools into a single data model. Explore it in the Keypup platform with the AI Agent, query it from your own AI assistant through the MCP server, or pull it into your systems with the API.
Keypup connects to the applications your teams already use, unifies their data into six datasets, and serves it through three products that share the same reporting engine.
Dashboards, insights, templates and the AI Agent, on top of your unified engineering data. Built for leaders and teams who report and decide inside Keypup.
The same data and reporting engine, inside the AI assistants your teams already use: Claude, ChatGPT, Cursor, VS Code, Kiro and any MCP client.
GraphQL access to the platform and the reporting engine, to feed your data warehouse, BI tools, portals and automations.
Every business question on our homepage maps to a template or an insight, built on a specific part of the data model. Start from the template, or simply ask the AI Agent.
The Value Stream Management dashboard template: workload categorisation, project effort distribution and planned versus unplanned work.
Ask the AI Agent: "Split of merged work between features, bugs and maintenance, per product, this quarter" 
Insights grouped by epic or by any Jira field, with progress, due dates and sprint commitment versus completion.
Ask the AI Agent: "Progress of every epic in our Q4 goals, with their due dates" 
Custom KPIs summing time spent and logged per epic or initiative, compared with the original estimates. Add your own rates with a formula.
Ask the AI Agent: "Hours logged per epic last quarter versus their original estimates" 
The DORA Metrics dashboard template. Lead time can start at the pull request or when work starts, measured from issue status changes.
Ask the AI Agent: "Lead time for changes per team, month by month, over the last year" 
Change failure rate templates, pull request or commit based, time to restore and bugs opened versus resolved, adapted to your own naming conventions.
Ask the AI Agent: "Bugs opened versus resolved per product over the last six months" 
AI usage templates: active users, suggestion acceptance and AI code contribution, placed next to pull request throughput and cycle time.
Ask the AI Agent: "Copilot active users next to PR cycle time, week by week" Everything you need to measure, explain and share engineering performance, in one place.
The AI Agent
The AI Agent works on your real data model. It builds and edits dashboards and insights from a conversation, explains what it did, analyzes the results and recommends what to do next.
Describe what you need and the agent creates the dashboard: insights, filters, layout and descriptions. Click Analyze and it comments on the results, flags bottlenecks and proposes the dashboard that will track your next improvement.
The agent picks the dataset, metrics, dimensions, filters and chart, writes the custom formulas, and explains the result in plain English. You review every change before it is saved.
AI Instructions give the agent persistent knowledge of your conventions: your workflow statuses, your labels, what counts as a bug or a release. Set them once: every conversation applies them, and AI assistants connected through the MCP server can read them too.
More about the AI Agent
Connect Claude, ChatGPT, Cursor, VS Code, Kiro or any MCP-compatible assistant to Keypup. Ask in plain language; the assistant discovers your data, builds the query, runs it on Keypup's reporting engine and explains the answer.
list_companiesFinds the teams you belong toget_company_ai_instructionsReads your team's AI Instructionslist_datasetsDiscovers the six datasetslist_dataset_fieldsLooks up the fields it needslist_formula_operatorsLearns the formula languagegenerate_dataset_queryTurns your question into a queryquery_datasetRuns it on the reporting engineYou ask one question. The assistant orchestrates the Keypup tools on its own: it finds your team, reads your AI Instructions, picks the dataset, generates the query and runs it. Then it explains the result and lets you refine it: "now break that down by repository".
list_companiesAcme Engineeringget_company_ai_instructionsConventions loadedgenerate_dataset_queryISSUES_PULL_REQUESTSquery_dataset3 rowsAverage PR cycle time last quarter, slowest first:
Most of the api-gateway delay is time waiting for a first review.
Illustrative conversation
Two GraphQL APIs give programmatic access to everything you see in Keypup, with the same datasets, formulas and permissions.
Every application is mapped onto the same model, so a GitHub pull request and a Jira issue can sit in the same report, linked together. Each field is documented in our Help Center.
Delivery, cycle time, progress, work types, due dates, sprints, estimates
Status changes, assignments, logged work, opened versus closed over time
Code volume: lines added, deleted and changed
Review behaviour and approval flow
Collaboration and discussion on issues, pull requests and reviews
AI adoption, acceptance and output, next to delivery
| Dataset | GitHub | GitLab | Bitbucket | Azure DevOps | Jira | Trello | ClickUp |
|---|---|---|---|---|---|---|---|
| Issues & pull requests | 85/88 | 83/88 | 63/88 | 87/88 | 63/88 | 58/88 | 61/88 |
| Activity events | 9/11 | 9/11 | 7/11 | 10/11 | 11/11 | 9/11 | 6/11 |
| Comments | 7/7 | 7/7 | 7/7 | 7/7 | 7/7 | 7/7 | 7/7 |
| Commits | 10/10 | 8/10 | 10/10 | 10/10 | – | – | – |
| Reviews | 10/10 | 10/10 | 10/10 | 10/10 | – | – | – |
| AI usage | Copilot | – | – | – | – | – | – |
Fields filled per dataset. Code hosts fill the code datasets, work trackers fill the planning fields such as story points, estimates and logged time; combine both for the full picture. AI usage also comes from Claude and OpenAI platforms, in beta.
Connect your tools and get your first dashboards in minutes, or let us walk you through the product.