unravel.Start free

Features / AI agents · MCP

Point your AI agent at production.

Connect Claude Code, or any MCP client, and it pulls the real context itself: the whole causal chain behind the failure, down to the query and the job. An alert fires, you ask “why is checkout failing?”, and the agent fixes it on the first try because it can see what actually happened.

Start freeRead the docs

Connect

One command, read-only access.

Run it in the repo you're debugging; grab the command with your key from Settings → API keys. Each key is pinned to one project and environment, shown once, and revocable. Available on Solo and up.

MCPConnect Claude Code
$claude mcp add --transport http unravel https://api.unravel.run/mcp

Read-only · scoped to one project & environment · the agent can look, never mutate. It sees every trace in your window, not a sample.

Search tracesRead a traceCompare two runsInspect exceptionsFind slow queriesRead the load

The exact tools, scope and limits are in the docs.

app.unravel.run/…/traces/019eb29d…
A production trace in Unravel: N+1 insights above the span timeline

The live context the agent reads: 75 events, two N+1 clusters, the SQL and the user's journey, interpreted into a story it can act on.

Or hand it over yourself

No MCP? Copy as Markdown.

Open any trace and press Copy as Markdown: the failing span, immediate cause, normalized queries, N+1 clusters and the user's journey, as text. Paste it into Claude Code, Cursor, or anything else.

incident.mdCopy as Markdown
## GET /shops/{id}/menu · 226 ms · 200
trace 019eb29d · production · 75 events
**N+1** · `categories` repeated 24× · 7.9 ms total
### Timeline
```
GET /shops/{id}/menu226 ms
├─ MenuController@show217.4 ms
│ ├─ select * from "menu_items" where "shop_id" = ?11.4 ms
│ └─ select * from "categories" where "id" = ?×247.9 ms
└─ view shop/menu38.4 ms
```
### User journey
- 14:31:48 · GET /shops/284 · 142 ms
- 14:31:55 · PUT /cart/items · 89 ms
⋯ 102 more lines→ Claude Code · Cursor

Keep exploring

The rest of the toolkit.

Async first

Async is where PHP apps break.

Websockets, queued jobs and scheduled work are traced like requests, with their own queries, failures and timing.

ExploreCausality

Causality, not log soup.

Normalized query fingerprints, collapsed N+1s and user journeys; every failed trace carries its immediate cause.

ExploreAlerting

Know before your users do.

Rules on new exceptions and error spikes, delivered to Slack, email or webhook with the debugging context inline.

Explore

Stop reconstructing incidents from logs.

Connect your real app in two minutes. Free forever on a 3-day window, the full debugger included.

Start free