Documentation

Stop AI agents from making irreversible mistakes

  • FINTECH — $340K wire to an OFAC-sanctioned shell company. Penalty: $500K+.
  • LEGAL — $2.4M contract signed with unlimited liability in section 14.3.
  • HEALTHCARE — Psychiatric records pulled with no treatment relationship. Fine: $1.5M.

One model. One guess. No review. consensus.tools makes multiple specialist agents evaluate every high-stakes decision before it executes.

Zero to Hero — 5 minutes

Build a guard system where 3 AI personas evaluate a customer service response before it sends. Legal reviewer catches unauthorized commitments. Compliance catches policy violations. CX analyst flags tone issues. The email gets blocked.

pnpm add @consensus-tools/core @consensus-tools/policies
Start the tutorial →

7 built-in guard types

send_emailUnauthorized pricing guarantees, legal commitments
code_mergeSQL injection, failing CI, missing tests
deploymentMissing rollback plan, schema incompatibility
support_replyUnauthorized refund promises, tone issues
agent_actionHigh-risk API calls, data mutations
permission_escalationOff-hours admin access, scope violations
publishPII exposure, blocked-word lists

Getting Started

See what goes wrong without consensus, then build a working guard system in 5 minutes.

  • → Overview
  • → Zero to Hero
  • → Install CLI
  • +3 more

CLI Reference

Complete reference for every CLI command — post jobs, submit, vote, resolve.

  • → Overview
  • → init
  • → config
  • +8 more

Engine

Deep dive into consensus policies, slashing economics, and security architecture.

  • → Overview
  • → Architecture
  • → Consensus Policies
  • +3 more

Packages

API reference for every package in the monorepo — from schemas to MCP.

  • → Overview
  • → schemas
  • → secrets
  • +11 more

Apps

Standalone applications — local board server, dashboard.

  • → Local Board

Examples

Fintech AML screening, healthcare HIPAA compliance, legaltech contract review — real regulatory scenarios.

  • → Fintech: Transaction Guard
  • → Healthcare: Data Access Guard
  • → Legaltech: Contract Review
  • +1 more

FAQ

Frequently asked questions and troubleshooting.

  • → FAQ

Jump to

→ Zero to Hero (5 min)→ Fintech: AML Screening→ Legaltech: Contract Review→ Healthcare: HIPAA Data Access→ Guards API→ 9 Consensus Policies→ Package Reference→ MCP (29 tools for Claude Code)