Rebase + your agents. Context over MCP.
Connect Claude, Cursor or any MCP-compatible agent to Rebase. Your tickets become structured context the agent can read and act on — right inside the tools your engineers already use.
Connect an agent
Point Claude Code, Cursor or any MCP-compatible client at your Rebase project.
Tickets become context
Every ticket is exposed as a structured, machine-readable payload — element, console, network and repro steps.
The fix starts at the source
The suspected source file gives agents a head start on where to look.
// The context layer
Data, not screenshots.
Agents can't read a screenshot pasted into Slack. Rebase hands them what they actually need: structured session context they can parse, reason about and act on.
- Five tools: list, search, get, comment, update status
- The screenshot travels too — as an image, up to 10 MB
- Local over stdio — the only traffic is HTTPS to the Rebase API
{ "mcpServers": { "rebase": { "command": "npx", "args": ["-y", "@rebasedotdev/mcp"] } }}// In the editor
Where the work happens.
No new dashboard. Your engineers ask their agent about a ticket from inside Claude Code or Cursor, and the agent pulls the context straight from Rebase.
- Claude Code and Cursor, out of the box
- Semantic search over 90 days — zero credits
- rbk_ tokens: scoped, shown once, stored hashed
// The details
What your agent can see. Per ticket, over MCP.
- The exact browser, OS and viewport
- Console errors and warnings, as they happened
- Failed and slow network requests
- A full-page screenshot from the moment it was flagged
- The exact clicks and inputs that led to the bug
- The flagged element, its selector and component name
- AI triage: a summary, a severity and the likely source
- A link back to the full Rebase report
Frequently asked questions
Connect Claude, Cursor, or any MCP-compatible agent to Rebase. Your tickets become structured context the agent can read and act on — right inside the tools your engineers already use.
Anything that speaks MCP — Claude Code and Cursor work out of the box. Five tools cover the loop: list, search, read, comment and update status.
With rbk_ tokens minted in the widget: explicit scopes (tickets:read, tickets:write, comments:write), shown once, stored only as a hash, 365-day default expiry, revocable any time. Read-only agents are a first-class setup.
No. We never train models on your data, and everything stays isolated to your tenant.
// More integrations
Works with the rest of your stack. One layer, every tool.
Start shipping feedback faster.
Start your 7-day free trial.