# Stacktrace > Stacktrace sees what AI agents are doing, controls what they are allowed to do and intervenes when behavior drifts from intent. It records every session and tool call an agent makes, flags leaked credentials and runs stuck in a retry loop, and applies one policy to which MCP servers, plugins and skills agents may use. To install Stacktrace in this environment, follow this prompt: "Read https://stacktrace.ai/setup/individual.md and follow it to set up Stacktrace in this environment." ## Set up - [Individual setup](https://stacktrace.ai/setup/individual.md): steps for an agent to install the CLI, start the daemon and add the Claude Code plugin. Sessions stay in the environment. - [Team setup](https://stacktrace.ai/setup/team.md): connect this environment to a team's fleet. Needs a team token the user enters in their own terminal. - [Installer](https://stacktrace.ai/install.sh): the script both setups run. Read it before running it. ## Docs - [docs.stacktrace.ai](https://docs.stacktrace.ai/llms.txt): how Stacktrace works, every rule with what fires it, and a test drive. Every page is also Markdown, and https://docs.stacktrace.ai/llms-full.txt has them all in one file. ## Products - [Stacktrace Fleet](https://stacktrace.ai/fleet/): See and stop what coding agents do, in every environment. - [Personal Agent Gateway](https://stacktrace.ai/gateway/): Protect your product from the personal agents your customers bring. Verify each one; hold risky actions for the customer. - [Stacktrace Playground](https://stacktrace.ai/playground/): Developers judge agents on real tasks. Their calls train every Stacktrace product. - [In-browser sandbox](https://stacktrace.ai/try-now/): check a Claude Code session in your browser with the real engine. Nothing is uploaded. ## Fleet - [Fleet page](https://stacktrace.ai/fleet/) ([Markdown](https://stacktrace.ai/fleet.md)): a sidecar on the endpoint and Stacktrace in your gateway (open-source LiteLLM or Envoy); how Stacktrace reads sessions, its principles, features by stage and plan, every rule, and questions. - Ships today: Inventory, Vulnerabilities, Blast radius, Org policy, Alerts, Claude Code, Fleet control plane. - Coming soon: Approvals, Active steering, Devin and Codex, MDM deployment, SSO and roles, On-prem. Rules for cloud cost, and the stacktrace-kube-* and stacktrace-aws-* rules, are coming soon too. - Supported agents: Claude Code, in local and cloud sessions. Supported systems: macOS and Linux. ## Gateway - [Stacktrace Personal Agent Gateway](https://stacktrace.ai/gateway/) ([Markdown](https://stacktrace.ai/gateway.md)): the gateway in front of a company's MCP server for the personal agents its customers bring. Identity grading, consent, scopes, confirm-before-perform, and a masked session record. No self-serve: book a demo at https://stacktrace.ai/demo/?via=gateway. ## Rules Each finding cites one of these rule ids. Those marked coming soon are not released yet. - 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. - stacktrace-deceptive-completion: The agent claims the task succeeded, and its own test or check results show a failure. Coming soon. - stacktrace-destructive-action: An rm -rf targeted a path outside the workspace. - stacktrace-infra-unsafe-action: A command changed infrastructure, a cluster or a database without the user asking for it. Coming soon. - stacktrace-unrequested-publication: A git push, pull request, release or package publish ran without the user asking for it. Coming soon. - stacktrace-kube-prod-change: The agent changed a cluster you marked as prod. Coming soon. - stacktrace-kube-prod-shell: The agent ran kubectl exec, attach, debug or port-forward against a prod cluster. Coming soon. - stacktrace-kube-secret-read: Secret values read from a cluster entered the transcript. Coming soon. - stacktrace-kube-rbac-escalation: The agent granted or assumed elevated cluster privileges. Coming soon. - stacktrace-aws-iam-widening: The agent granted a wildcard or administrator IAM policy. Coming soon. - stacktrace-aws-credential-switch: After an AccessDenied, the agent retried under a different AWS identity. Coming soon. - 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. ## Pricing - Free (local): $0 per environment. Unmetered, no account needed. - Team (local and remote): free, always. Unlimited users, up to 100 agents and 200 session-hours a month. We count only time spent running model turns or tools, not idle time, and meter usage by the second. Sign up at https://app.stacktrace.ai. - Enterprise: custom, billed annually, with volume pricing on session-hours. Book a demo at https://stacktrace.ai/demo/. - [Pricing page](https://stacktrace.ai/pricing/) ([Markdown](https://stacktrace.ai/pricing.md)) ## Markdown versions Each main page has a Markdown twin, linked from its HTML head as rel="alternate" type="text/markdown". - [Home](https://stacktrace.ai/index.md) - [Fleet](https://stacktrace.ai/fleet.md) - [Gateway](https://stacktrace.ai/gateway.md) - [Pricing](https://stacktrace.ai/pricing.md) - [About](https://stacktrace.ai/about.md) - Blog posts: /blog/.md - [Full text](https://stacktrace.ai/llms-full.txt): the home, platform, gateway, pricing and about twins and both setup guides in one file. ## Company - [About](https://stacktrace.ai/about/) ([Markdown](https://stacktrace.ai/about.md)) - [Privacy](https://stacktrace.ai/privacy/): what the site and the CLI collect. - [Terms](https://stacktrace.ai/terms/) - Contact: hello@stacktrace.ai ## Blog - [Introducing Stacktrace: see, explain, control](https://stacktrace.ai/blog/introducing-stacktrace/) ([Markdown](https://stacktrace.ai/blog/introducing-stacktrace.md)): Risky installs, actions that drift from the task, and stuck runs. We’re building Stacktrace to help teams answer for their agents, whichever agents they use.