BETAATP 1.1

Building the standard for autonomous agents.

Agent Transaction Protocol (ATP) is the middleware for the agentic economy. It provides the audits, permissions, and human-in-the-loop controls needed to trust autonomous work.

Without governance, the agentic economy is chaos.

Agents are already acting autonomously: sending messages, committing transactions, calling external systems. There is no shared standard for what should happen before any of that executes. No interception. No receipt. No recourse.

ATP is the answer to that: a protocol layer that brings order. Every action intercepted, evaluated, and receipted before anything touches a downstream system.

No interception

An agent commits a payment to an external party. No approval was requested. No record exists.

No audit trail

A compliance audit requests a log of agent decisions. There isn't one. Actions were fire-and-forget.

No oversight

An agent crosses an organisational boundary mid-task. No human was notified. No halt was triggered.

No recourse

Something goes wrong. There is no receipt, no rule reference, no way to trace what was authorised.

Agent Transaction Protocol

A standard for what happens when an agent acts.

ATP defines the contract between an agent runtime and the systems it touches: interception, evaluation, decision, and a signed receipt.

01

Interception

Every agent action passes through the ATP checkpoint before it reaches any downstream system.

02

Permission model

Allow, deny, or escalate to human approval. Configurable per action type, topology, and risk level.

03

Trust receipts

Each decision produces a signed, immutable receipt. Outcome, rule, topology, and halt condition all included.

04

Audit trail

A complete, tamper-evident log of every agent action and its governance decision, ready for compliance review.

Coming soon

Transient Trace

A runtime-agnostic governance layer, built to the ATP standard. Trace works regardless of which agent framework you use. It intercepts, evaluates, and receipts actions without being tied to any specific runtime.

Action monitoringReal-time visibility into what agents are doing and why.
Approval workflowsHuman-in-the-loop gates for high-risk or cross-boundary actions.
Runtime agnostic layer.Connect to any Agent Runtime stack.
Compliance exportTrust receipts and audit logs ready for review.

Works with any agent framework

LangChain
AutoGen
OpenAI Agents SDK
CrewAI
Manus
HermesAgent
OpenClaw
OpenAI Frontier
Claude Dispatch
Google ADK

If it emits agent actions, Trace can govern them.

Confidence, control and compliance by design.

Configurable Data Burn

Lowers legal and compliance risk, and makes procurement easier.

Cloud Armor Protection

Better uptime and fewer incidents that interrupt workflows.

API Hardening and Access Control

Reduces unauthorised access risk and protects sensitive data before the data is burned.

Operational Visibility

Faster incident detection and resolution, less downtime, and more reliability.

Start with the protocol
Read the published standard, then use internal tools when you need API or MCP access.
ATP 1.1 standardGet API key (free)

No credit card required • 50 free credits included