stacktrace

Backed by Combinator

Try Stacktrace
FleetSign up. Every agent's findings in one place. Sandbox (in-browser)Check a session in your browser.

Stacktrace Fleet

Control the coding agents your team runs.

See what every coding agent does, in every environment, and stop the unsafe actions as they happen.

Try now

See.

Know what actually ran.

See each tool call your agents make, what came back, and which skill, MCP server, plugin or hook made it.

Stacktrace knows a component wherever it runs. When one causes trouble, you see every other environment and agent that uses it.

Stacktrace Fleet Findings: ten findings across six agents, with a credential sent to an MCP server open, showing the evidence, the fix and the three agents it ran on

Control.

Set the limits ahead of time.

One policy for every agent: allow a component, block it, or require approval before a consequential action.

Write it once and enforce it across local and remote agents.

Stacktrace Fleet Policy: MCP servers allowed except one blocked, with the saved revision and who saved it

Intervene.

Real-time alerts.

A finding reaches Slack or Claude Code while the agent is still running, with the evidence and the fix.

Coming soon Steer agents dynamically.

A Stacktrace message in the #agent-alerts channel: "A service refused the request, and the agent did not get what it asked for", with the subject, evidence, severity, confidence and session

On the endpoint, at the gateway, or both.

Sidecar

Endpoint

A sidecar process beside each coding agent, on a laptop, CI runner or cloud session. Read-only, never in the agent's path.

Open-source LiteLLM · Envoy

Gateway

Runs inside the gateway you already operate, and stops a dangerous call before the agent runs it.

How it works.

ON THE ENDPOINT · LAPTOP, CI RUNNER OR CLOUD SESSION Claude Codethe agent Session filesJSONL + MCP logs Stacktrace sidecar 1 · Read sessions 2 · Inventory match 3 · Deterministic rules policy read-only Local consoleevery finding alerts transcripts and tool output stay here AT THE GATEWAY · SUCH AS LITELLM OR ENVOY Your agentsthrough the gateway Stacktrace, inline 1 · Check each tool call 2 · Block, or pass it on calls blocked Findings logevery block, for audit blocked calls never reach the agent each box names what it receives MODEL PROVIDER Anthropic, OpenAI, …the calls your agentsalready make calls responses, checked before the agent STACKTRACE FLEET Fleetfindings and inventory, all agents Slack Org policyallow and block lists findings, inventory policy, compiled locally OTHER SERVICES Webhooknotifying findings, if configured OSVpackage names and versions usage telemetry: counts, no session content

Sessions stay in your environment.

Fleet gets findings and inventory, never transcripts or tool output.

Precision before recall.

A rule ships only once its findings get acted on.

Out of the agent's path, unless you put it there.

The sidecar reads sessions read-only. Only the gateway sits inline, so it can stop a call.

Stacktrace rules.

Every finding cites its rule, the evidence and a fix. Every rule, with what fires it →

Reliability

  • stacktrace-tool-retry-storm The same tool call, with identical arguments, failed three or more times in a row
  • stacktrace-blocked-policy The auto-mode safety check refused an action and will refuse that kind for the rest of the session
  • stacktrace-blocked-throttled The model provider throttled requests with a 429 or 529, and the agent may not return to the lost work
  • stacktrace-blocked-upstream A service refused a request with a 401 or 403, and the agent moved on without the result
  • stacktrace-blocked-quota The account hit its spend or usage limit, and model calls stop until it resets

Operations

  • stacktrace-destructive-action An rm -rf targeted a path outside the workspace

Security

  • stacktrace-credential-egress A credential-shaped value appeared in the arguments of an outbound tool call
  • stacktrace-credential-persisted The agent wrote a secret-shaped value to a file
  • stacktrace-vulnerable-dep-invoked A tool call reached a component with a known advisory
  • stacktrace-mcp-transient-failure An MCP server failed authentication, then a later call to it succeeded

Questions.

What is Stacktrace?

Stacktrace reads the session files AI coding agents write and checks each session against detection rules. Findings go to a local console, to Slack and into the running Claude Code session. On Team, Fleet shows the findings and inventory of every agent an org runs and serves one policy to all of them.

Which agents does Stacktrace support?

Claude Code, in local sessions and in cloud sessions. Devin and Codex are in preview.

Does Stacktrace upload my transcripts?

No. Transcripts and tool output are read in the environment where the agent runs, and Fleet receives findings and inventory with no conversation text.

What does Stacktrace detect?

Stacktrace checks sessions against 10 rules in the reliability, operations and security families, listed above. Examples are stacktrace-credential-egress for a credential in an outbound tool call and stacktrace-tool-retry-storm for a tool call that failed three or more times in a row. The stacktrace-kube-* and stacktrace-aws-* rules and the cost rules are coming soon.

How do I install Stacktrace?

Paste this prompt into Claude Code: "Read https://stacktrace.ai/setup/individual.md and follow it to set up Stacktrace in this environment." The agent runs the installer, starts the daemon and adds the Claude Code plugin.

How much does Stacktrace cost?

The local sidecar is free, with no account. Team is free too, always: unlimited users, up to 100 agents and 200 session-hours a month. Beyond that, Enterprise is custom and billed annually; book a demo at stacktrace.ai/demo.

Start with your next session.

Free for your team. Findings in Slack or Claude Code while the agent is still running.

Try now
talk to the founders