CA: launching soon
004 — Built for any agent

One policy layer.
Every kind of agent.

Checkpoint sits in front of any action an agent can take. The rules are yours; the verdict comes back before anything runs. Here is what that looks like across six common settings. Play any scenario to watch the check happen.

01 — FINANCE

Agents that move money

An agent paying invoices on its own. Payments under your limit clear instantly; anything larger stops for human approval.
if amount > $25,000 → ESCALATE to finance
elseALLOW
02 — DEVELOPERS

Coding agents that ship

A coding agent working overnight. Routine changes merge; production deploys escalate, and secrets stay off-limits.
if env == "production"ESCALATE to eng-lead
if touches secrets → STOP
03 — INTERNAL OPERATIONS

Assistants with database access

An assistant updating records company-wide. Routine edits run; mass changes pause for review before they touch thousands of rows.
if rows > 5,000 → STOP
elseALLOW
04 — ON-CHAIN AGENTS

Agents that sign transactions

An agent moving tokens and calling contracts. Small transfers clear; large ones and contract approvals route to a human first.
if usd > $50,000 → ESCALATE
if contract approval → ESCALATE
05 — DEVOPS & INFRASTRUCTURE

Agents that manage infrastructure

An agent scaling and tearing down resources. Routine ops run; destructive actions on production are blocked cold.
if destructive and prod → STOP
elseALLOW
06 — TRADING & TREASURY

Agents that execute trades

An agent executing trades and moving funds. Trades within mandate clear instantly; anything over threshold routes to a human.
if notional > mandate → ESCALATE
elseALLOW

Your agent, your rules.

Wrap any action in one line. Define the limits that matter to you.