FailproofAI provides local observability, session history, audits, and policy enforcement for AI agent runs across supported agent harnesses. It hooks supported harnesses to capture runs and block dangerous tool calls before execution.
Project overview
Its tool-call policies block dangerous calls before they run across all twelve supported harnesses, and its local dashboard reads run history on the machine without an account, signup, or data leaving the box.
Project type
AI Agent · Workflow
Use cases
Automation
Deployment
Refer to project documentation
License
License pending
Best for
Developers and AI engineers who need runtime observability and enforcement for AI agent runs in the twelve supported harnesses.
Users who want a local dashboard that reads run history on the machine without an account or signup.
Key capabilities
Hooks supported agent harnesses to capture runs and block dangerous tool calls before execution.
Serves a local dashboard showing sessions, model calls, tool calls, hook decisions, blocked actions, and policy guidance.
Scans local run history for risky patterns and suggests policies to stop them.
Provides tracing, sessions, and audits for agents that do not run in a supported harness; enforcement requires a hook in the agent runtime.
Limitations and risks
For agents outside the supported harnesses, the Python SDK provides tracing, sessions, and audits, but enforcement requires a hook in the agent's own runtime.
Enforcement capability varies by harness: tool-call blocking is verified on all twelve, while turn-end gates are verified on eight.
The sanitize-* policy family runs after a tool returns, so it reports secrets in tool output rather than keeping them out of the context.
Getting started
Install globally with npm, run configuration, add the policy set, and start the CLI dashboard: npm install -g failproofai; failproofai config; failproofai policies add FailproofAI/policies; failproofai. Setup involves a global npm install, configuration, policy selection, and starting the CLI dashboard.
Evidence and sources
GitHub project description: Observability and enforcement for AI agent harnesses. Capture every run and runtime reliability with policy enforcement.
README: Twelve harnesses in two classes — ten coding CLIs, and two chat and assistant gateways (Hermes, OpenClaw). One policy API and one session history across all of them. What a policy…
Release: failproofai 1.0.9
README: Every one of these gates the call *before* it runs, so they hold on all twelve harnesses.
README: Run `failproofai` with no arguments and it serves a dashboard on `localhost:8020` reading the run history already on your machine — no account, no signup, nothing leaving the box.…