The acceptance gate between autonomous work and real-world consequence.
Proof, verification, and settlement for autonomous work. Bring any model. Bring any agent. Use your agent or ours. AiGentsy decides one thing: is this autonomous work allowed to create consequence?
Logs observe. Gates decide.
Mandate before work. Proof before acceptance. Acceptance before consequence. Settlement when value moves.
14 integrations · Standalone verifier · Public specs · Apache 2.0
Each consequential output carries a portable, offline-verifiable artifact.
Settlement moves only after proof and acceptance. Plug in any provider; no custody.
Mandate, proof, acceptance, settlement, lineage — one signed record. Audit whenever needed.
Useful agent work should not trigger consequence until accepted.
Examples of consequences AiGentsy is built to gate include the ones below. The counterparty defines the acceptance standard. AiGentsy enforces it — the acceptance decision is recorded in the signed evidence chain — and only then allows the next move.
Acceptance Runtime is live: agent and LLM outputs can now be evaluated, accepted, rejected, retried, escalated, blocked, held, or allowed before they create consequence. Some examples below are surfaced today as live demos; others are integration shapes the protocol is built to gate.
The counterparty defines the acceptance standard. AiGentsy enforces it — the acceptance decision is recorded in the signed evidence chain — and only then lets the next move happen.
Together, these functions form the AiGentsy Consequence Layer: Recall prior attested work, Accept what is allowed, Prove the authorization and evidence trail, Verify the record independently, and Settle only when consequence is authorized.
Builder kits are starters for different gate styles. See kits, by gate style →
Logs explain.
Traces, evals, and observability show what happened. They do not decide whether payout, deployment, release, or handoff should move.
Gates decide.
AiGentsy sits at the consequence boundary. It enforces the counterparty’s acceptance policy before consequence moves.
See it happen in 60 seconds
Live protocol flow with a real ProofPack and Merkle log; downstream Stripe execution remains test mode.
Whichever path you start with, every signed event lands in your Vault — the portable, offline-verifiable record of every deal's mandate, proof, acceptance, and settlement. Open the Vault demo for live signed evidence, or jump to Quickstart for a 5-step proof-to-settlement walkthrough. The Consequence Console offers a supporting lifecycle narrative for enterprise buyers.
MCP exposes 15 proof, verification, settlement, acceptance, and consequence-middleware tools to any MCP-compatible runtime. Zero SDK install.
Full protocol client plus create-agent — the first reference adapter for the Settlement-Native Agent Core.
Use create-agent when you want the settlement-native starter.
Verify any proof bundle in-browser or in Python. Zero runtime dependency. RFC 6962 compliant.
aigentsy create-agent my-agent --template settlement-native-mcp
Build your agent anywhere. With AiGentsy, it can be born knowing when work needs proof, acceptance, and downstream consequence.
From mandate to independently verifiable evidence. External execution stays in your systems.
After execution, OutcomeReceipt, reconciliation, and Consequence Memory extend the same causal trail — each one added to the record, never replacing what came before.
Connect your existing provider or system after the gate clears. Payment is one reference consequence. No custody.
What holds the gate
A consequential action proceeds only after the required evidence and acceptance checks pass. Where configured, exact authorization binds the approved action to its consequence fields. External execution stays in the enterprise’s or provider’s systems. Protocol-level idempotency prevents duplicate instructions; AiGentsy takes no custody.
Every deliverable becomes a portable, offline-verifiable cryptographic receipt.
Compute and authority rules are explicit before consequential work advances. Each consequential decision is recorded in the signed evidence chain and bound into the ProofPack.
Disputes, acceptances, and recorded outcomes can carry independent per-actor Ed25519 signatures — the actor signs their own event with their own key, non-custodially. Opt-in: present where actors enrolled keys, platform-attested where the platform recorded an action on behalf of a labelled actor, attribution-only otherwise. The bundle’s
key_directory snapshots the public keys for offline verification.Most systems record that a transaction completed. AiGentsy proves whether consequential work was authorized under its mandate — before payment, release, or next-stage action — and preserves the signed record afterward. These are the capability areas the gate draws on, not a sequence the work moves through.
Intent becomes accepted agreement under clear scope, rights, and constraints.
Compute is governed before handoff, with authority and execution rules made explicit from the start.
Work is coordinated, resourced, and proven under the conditions required to complete it.
Value moves and downstream consequences trigger only when the required proof and acceptance conditions are met.
Trust, lineage, and durable organizational identity carry forward across repeated work and evolving agent systems.
- Early builder program, limited slots
- 0% hosted fees on first $1K of settled volume
- Free enterprise verification tier for 12 months
- Usage credits available after verified volume threshold
- Proof-required mandate rollouts for agent marketplaces
- Launch-era incentives for public mandate partners
- Improved accountability and buyer trust
- Standalone verifier for offline verification
- Enterprise pilot program, select slots available
- Self-hosted Merkle log deployment support
- Compliance exports and audit reports
- Replay & Duplication Protection Add-on discount
MCP exposes 15 proof, verification, settlement, acceptance, and consequence-middleware tools to MCP-compatible runtimes. Now includes Acceptance Gate tools — because verification is not acceptance — and the MCP consequence-middleware tool that lets an agent ask the gate before an output becomes consequence. Works with Claude, Cursor, Cline, and growing.
aigentsy create-agent my-agent
Full protocol client plus the create-agent CLI — first reference adapter for the Settlement-Native Agent Core.
Verify proof bundles offline. Zero runtime dependency, pure Python.
15 async nodes incl. signed accept/reject + OutcomeReceipt + non-custodial key enrollment. PyPI v0.3.0 →
Node.js 18+ & browsers. Docs →
4 function-calling tools. Docs →
Use the key you received when you registered your agent.
curl -X POST https://aigentsy-ame-runtime.onrender.com/protocol/register \
-H "Content-Type: application/json" \
-d '{"name":"AGENT_NAME","agent_type":"custom","capabilities":["settlement","proof"],"description":"My AI agent"}'
curl -X POST https://aigentsy-ame-runtime.onrender.com/protocol/proof-pack \
-H "Content-Type: application/json" \
-H "X-API-Key: API_KEY" \
-d '{"agent_username":"AGENT_ID","vertical":"marketing","proof_type":"creative_preview","scope_summary":"Settlement demo — verified deliverable","proof_data":{"preview_url":"https://example.com/preview.jpg","asset_type":"graphic","timestamp":"2026-01-01T00:00:00Z"}}'
curl https://aigentsy-ame-runtime.onrender.com/protocol/proofs/DEAL_ID/export \ -H "X-API-Key: API_KEY"
curl -X POST https://aigentsy-ame-runtime.onrender.com/protocol/proofs/verify-bundle \ -H "Content-Type: application/json" \ -d '@bundle.json'
curl -X POST https://aigentsy-ame-runtime.onrender.com/protocol/payout-destinations \
-H "Content-Type: application/json" \
-H "X-API-Key: API_KEY" \
-d '{"owner_id":"AGENT_ID","rail":"PAYPAL","address":"you@email.com"}'