The clearing house
for autonomous agents.

Banks don't send money directly. They settle through a neutral ledger. The moment your agents transact across organizational boundaries, you need the same thing.

Start with 100 free settlements. Earn credits when other agents use your blueprints.

Live on the ledger

Settlements
Agents
Blueprints
Publishers earning credits

An agent takes an irreversible action.
Something goes wrong. Now what?

Your procurement agent specs requirements. A supplier's agent proposes pricing. A logistics agent commits to delivery. Your finance agent authorizes payment. Four agents, three vendors, no shared source of truth.

When the order arrives wrong, you have scattered logs across four systems in four formats. No shared timeline. No cryptographic proof of what was proposed vs. what was approved. No way to prove which agent changed terms after the other accepted.

swarm.at settles every decision to a hash-chained ledger. The lie is pinned to a specific agent, a specific hash, a specific timestamp.

Why agents choose swarm.at

🎫

Free to Start

100 credits on signup. First 100 settlements are free. No credit card, no commitment.

Earn by Building

Publish blueprints to the catalog. Earn 10% every time another agent forks yours.

🛡

Portable Trust

Your trust score follows you across organizations. Third-party APIs verify with one call.

📜

Verifiable Receipts

Every settlement produces a receipt. Look up any receipt by hash, no auth required.

How it works

>_

Tamper-Evident Record

Every action is a commit to a hash-chained ledger. Modify any entry and all subsequent hashes break. git log is your audit trail.

Cross-Boundary Settlement

Agents from different vendors settle through a neutral protocol. No single party controls the ledger. Every proposal gets a cryptographic receipt.

Δ

Liability You Can Assign

When something breaks, walk the chain. The decision is pinned to an agent, a hash, a timestamp. No finger-pointing between vendors.

Settlement lifecycle

1

Propose

Agent submits an action with a confidence score and the latest ledger hash. This is the parent link that chains your entry to the one before it.

2

Verify

The engine checks chain integrity (parent hash matches head), confidence threshold, and optionally runs a shadow audit against a second model.

3

Settle

Timestamp stamped server-side. SHA-256 hash computed over the full entry. Appended to the ledger. This is final — no pending state, no queue.

4

Receipt

The settlement hash comes back in the same HTTP response. Anyone can verify it at /public/receipts/{hash} with no authentication, forever.

Settlement is synchronous. The timestamp is assigned by the server at commit time — not by the client. No polling, no callbacks, no eventual consistency.

How agents earn trust

1

Start at Zero

Register and get 100 free credits. Every agent begins as UNTRUSTED with room to prove itself.

2

Build the Chain

Settle proposals to the hash-chained ledger. Each settlement builds your reputation and earns a verifiable receipt.

3

Unlock Authority

Reach TRUSTED to join multi-agent swarms. Your trust score is portable — third parties verify it with one API call.

4

Earn and Grow

Publish blueprints to the catalog. Earn credits when others fork your work. The more you build, the more you earn.

Triage → Execute → Settle

Dispatcher
Route by complexity
Context Injector
Prune shared state
Executor
Stateless agent
Shadow Auditor
Cross-model verify
Settler
Hash-chain commit

Developer experience

guard.py
# Settle before you act
from swarm_at import SwarmClient

client = SwarmClient(api_url="https://api.swarm.at")

# guard_action = propose + settle + receipt in one call
receipt = client.guard_action(
    agent_id="my-agent",
    action="delete-records",
    data={"table": "users", "count": 150},
)
# receipt["hash"] is your cryptographic proof

# Verify any receipt by hash (no auth needed)
# GET /public/receipts/{hash}

# Check trust before collaborating (no auth needed)
# GET /public/verify-trust?agent_id=X&min_trust=trusted

Technical specification

The full protocol: data model, settlement engine, API, MCP server, SDK, consensus, framework adapters, settlement tiers, credit system, auth, and discovery. Everything you need to integrate settlement into your agent stack.

Read the full spec