# Failproof AI > Trace agent behavior, find failures with audits, and prevent repeat failures with deployed policies. - [Integrations and reference (120 pages)](https://docs.befailproof.ai/_llms/en/integrations-and-reference.md): Documentation for Integrations and reference. ## Start ### Start here - [Make your agent failproof](https://docs.befailproof.ai/index.md): Observability and enforcement for every harness your agents run in — coding CLIs, chat gateways, self-hosted assistants, and your own instrumented agents. - [Quickstart](https://docs.befailproof.ai/start/quickstart.md): Capture an agent session, find a failure, and start preventing it. - [Agent harnesses](https://docs.befailproof.ai/reference/harnesses.md): Capture sessions and enforce policies across all 12 supported agent harnesses. - [Failproof Assistant](https://docs.befailproof.ai/sessions/assistant.md): Analyze and operate Failproof AI in natural language, from questions and queries to dashboards and audits. - [Run your first failure check](https://docs.befailproof.ai/start/first-audit.md): Create an audit from the Cloud dashboard or the fp CLI and review its first findings. - [Prevent your first failure with a policy](https://docs.befailproof.ai/start/first-policy.md): Author a policy version, deploy it in observe mode, and then enforce it. - [Choose your setup](https://docs.befailproof.ai/start/setup.md): Choose local enforcement, Failproof AI Cloud, or an enterprise deployment. - [Core concepts](https://docs.befailproof.ai/start/concepts.md): The small set of concepts used throughout Failproof AI. #### Starter templates - [Custom agents](https://docs.befailproof.ai/start/quickstarts/custom-agents.md): Wrap your agent in three `with` blocks and it starts recording. - [LangChain and LangGraph](https://docs.befailproof.ai/start/quickstarts/langchain.md): Install, instrument, and see your first trace. - [CrewAI](https://docs.befailproof.ai/start/quickstarts/crewai.md): Install, instrument, and see your first trace. - [LlamaIndex](https://docs.befailproof.ai/start/quickstarts/llamaindex.md): Install, instrument, and see your first trace. - [Pydantic AI](https://docs.befailproof.ai/start/quickstarts/pydantic-ai.md): Install, instrument, and see your first trace. ## Trace Agents ### Understand every run - [Sessions](https://docs.befailproof.ai/sessions/overview.md): Start with the complete record of one agent run. - [Live events](https://docs.befailproof.ai/sessions/live-events.md): Watch agent activity arrive while a session is running. - [Read a trace](https://docs.befailproof.ai/sessions/read-a-trace.md): Find the event that changed the course of an agent session. - [Models](https://docs.befailproof.ai/sessions/models.md): Compare model latency, tokens, context use, and traffic distribution. - [Hooks](https://docs.befailproof.ai/sessions/hooks.md): Measure hook activity, triggers, and latency separately from policy outcomes. - [Policy decisions](https://docs.befailproof.ai/sessions/policy-decisions.md): See what policies evaluated, blocked, instructed, or allowed. - [Tools](https://docs.befailproof.ai/sessions/tools.md): Find slow, failing, or overused tools across agent sessions. - [Errors](https://docs.befailproof.ai/sessions/errors.md): Group repeated errors and open the sessions behind them. - [Online evaluations](https://docs.befailproof.ai/sessions/evaluations.md): Score live and completed sessions for quality, compliance, cost, and latency. - [Metrics](https://docs.befailproof.ai/sessions/metrics.md): Plot magnitude metrics such as latency, cost, and token use. - [Dashboards](https://docs.befailproof.ai/sessions/dashboards.md): Track the reliability signals that matter to an agent or workflow. - [Queries](https://docs.befailproof.ai/sessions/queries.md): Explore session, event, and evaluation data with reusable SQL. ### Plug in your agent - [Custom agents](https://docs.befailproof.ai/start/integrations/custom-agents.md): Instrument an agent you wrote yourself, or a framework without an adapter. - [LangChain and LangGraph](https://docs.befailproof.ai/start/integrations/langchain.md): Instrument graphs, nodes, tools, retrievers, and model calls with one call. - [CrewAI](https://docs.befailproof.ai/start/integrations/crewai.md): Instrument crews, flows, agents by role, tools, memory, and human feedback. - [LlamaIndex](https://docs.befailproof.ai/start/integrations/llamaindex.md): Instrument workflows, steps, function agents, and retrievers. - [Pydantic AI](https://docs.befailproof.ai/start/integrations/pydantic-ai.md): Instrument typed agents, tools, model calls, and retries. ## Find failures ### Find and manage failures - [Audits](https://docs.befailproof.ai/audits/overview.md): Review a defined population of sessions for failures that traces alone will not reveal. - [Audit local agent history](https://docs.befailproof.ai/audits/local-audit.md): Scan supported agent CLI histories offline and review risky or wasteful behavior locally. - [Set up an audit](https://docs.befailproof.ai/audits/setup.md): Define an audit goal, session population, and evidence context. - [Agent context](https://docs.befailproof.ai/audits/agent-contracts.md): Tell audits what each agent must do, produce, and never do. - [Run and review an audit](https://docs.befailproof.ai/audits/run.md): Run an audit, verify its coverage, and inspect the resulting findings. - [Audit cadence](https://docs.befailproof.ai/audits/cadence.md): Choose when recurring audits run and how much data they review. - [Findings and issues](https://docs.befailproof.ai/audits/findings-and-issues.md): Turn audit evidence into owned, trackable remediation work. - [Alerts](https://docs.befailproof.ai/audits/alerts.md): Detect recurrence and route an incident to the right responders. - [Audit recipes](https://docs.befailproof.ai/audits/recipes.md): Starting goals for common agent failure investigations. ## Prevent failures ### Prevent repeat failures - [Policies](https://docs.befailproof.ai/policies/overview.md): Observe, guide, or block agent actions before a known failure repeats. - [Builtin policies](https://docs.befailproof.ai/policies/builtin.md): Enable maintained guardrails for common agent failure modes. - [Builtin policy catalog](https://docs.befailproof.ai/policies/builtin-catalog.md): Review every builtin Failproof AI policy, its trigger, recommended state, and configurable parameters. - [Custom policies](https://docs.befailproof.ai/policies/custom.md): Write a policy for a failure mode unique to your agent workflow. - [Policy packs](https://docs.befailproof.ai/policies/packs.md): Install a set of policies published as a GitHub release, and manage what it enforces. - [Publish a pack](https://docs.befailproof.ai/policies/publish-a-pack.md): Ship your own policies as a GitHub release that anyone can install. - [Local configuration](https://docs.befailproof.ai/policies/local-configuration.md): Control policy scope, parameters, custom files, and machine-level Failproof AI settings. - [Policy editor](https://docs.befailproof.ai/policies/editor.md): Create and revise versioned policies from a confirmed failure mode. - [Deploy policies](https://docs.befailproof.ai/policies/deploy.md): Roll out a reviewed policy version to the intended machines. - [Deploy policies to machines](https://docs.befailproof.ai/policies/fleet.md): Know which machines are enrolled, current, and enforcing the intended policy versions. - [Rollback](https://docs.befailproof.ai/policies/rollback.md): Restore a known policy deployment when a rollout disrupts valid agent work. - [Failure behavior](https://docs.befailproof.ai/policies/failure-behavior.md): Understand what happens when policy evaluation or the local daemon is unavailable. ## Admin ### Operate Failproof AI - [Administration](https://docs.befailproof.ai/admin/overview.md): Operate access, usage, organizations, and security without mixing them into the reliability workflow. - [Usage](https://docs.befailproof.ai/admin/usage.md): Inspect organization consumption and the active billing window. - [Keys and permissions](https://docs.befailproof.ai/admin/keys-and-permissions.md): Create scoped API keys for machines, automation, and operators. - [Users and organizations](https://docs.befailproof.ai/admin/users-and-organizations.md): Control membership and keep each organization's data and actions scoped. - [Settings and security](https://docs.befailproof.ai/admin/settings-and-security.md): Configure operational settings and make deliberate choices about agent data. ## OpenAPI Specs - [openapi](/reference/openapi.json) ## Optional - [Support](mailto:support@befailproof.ai) > The links below point to documentation indexes. Follow each `/_llms/` index recursively until you reach documentation pages. ## Indexes - [English / Integrations and reference (120 pages)](https://docs.befailproof.ai/_llms/en/integrations-and-reference.md): Documentation for English / Integrations and reference. - [Chinese (67 pages)](https://docs.befailproof.ai/_llms/zh.md): Documentation for Chinese. - [Japanese (67 pages)](https://docs.befailproof.ai/_llms/ja.md): Documentation for Japanese. - [Korean (67 pages)](https://docs.befailproof.ai/_llms/ko.md): Documentation for Korean. - [Spanish (67 pages)](https://docs.befailproof.ai/_llms/es.md): Documentation for Spanish. - [Brazilian Portuguese (67 pages)](https://docs.befailproof.ai/_llms/pt-br.md): Documentation for Brazilian Portuguese. - [German (67 pages)](https://docs.befailproof.ai/_llms/de.md): Documentation for German. - [French (67 pages)](https://docs.befailproof.ai/_llms/fr.md): Documentation for French. - [Russian (67 pages)](https://docs.befailproof.ai/_llms/ru.md): Documentation for Russian. - [Hindi (67 pages)](https://docs.befailproof.ai/_llms/hi.md): Documentation for Hindi. - [Turkish (67 pages)](https://docs.befailproof.ai/_llms/tr.md): Documentation for Turkish. - [Vietnamese (67 pages)](https://docs.befailproof.ai/_llms/vi.md): Documentation for Vietnamese. - [Italian (67 pages)](https://docs.befailproof.ai/_llms/it.md): Documentation for Italian. - [Arabic (67 pages)](https://docs.befailproof.ai/_llms/ar.md): Documentation for Arabic. - [Hebrew (67 pages)](https://docs.befailproof.ai/_llms/he.md): Documentation for Hebrew.