OPEN SOURCE · MIT LICENSE

Your AI agents need
a permanent record.

Tamper-proof verification for every agent action. Your clients see exactly what happened — and can verify it independently.

Open protocol. Self-hostable. Once clients expect a treeship.dev/verify/ link, that expectation becomes the standard.

pip install treeship-sdk
View on GitHubRead the docs
3
lines to add
MIT
licensed
Self-host
or use hosted
Ed25519
signatures
treeship.dev/verify/my-workspace
LIVE
AgentActionResultStatus
How it works

One API call. Permanent verification.
No infrastructure changes.

agent.py
# Your agent logic stays exactly the same
from treeship_sdk import Treeship
ts = Treeship()
result = ts.attest(
agent="loan-processor",
action="Approved application #12345",
inputs_hash=ts.hash(application_data)
)
print(result.verify_url)
# → https://treeship.dev/verify/loan-processor/abc123

Add ts.attest() after any action. Treeship signs it with Ed25519 and creates a permanent verification URL.

1
Install the SDK
pip install treeship-sdk — or self-host the entire stack.
2
Call ts.attest() after actions
Pass agent name, action description, and a hash of inputs.
3
Share the verification URL
treeship.dev/verify/your-agent — permanent, public, no login required.
4
Independent verification
Anyone can verify signatures with just the public key. No trust required.
Try it live

See verification in action

Create a real attestation instantly — no signup required.

Try it now

Click the button to create a real, cryptographically signed attestation. You'll see the verification page instantly.

Creates a real Ed25519-signed attestation under the demo-agent

Or deploy your own agentView source
1. Clone and run
git clone github.com/zerkerlabs/treeship
cd treeship/examples/demo-agent
pip install -r requirements.txt
python agent.py
2. Create an attestation
curl http://localhost:8000/process \
-H "Content-Type: application/json" \
-d '{"applicant": "Jane", "amount": 50000}'
Integrations

Works with your stack

Drop-in integrations for popular AI agent frameworks.

C
Claude Code
Anthropic's coding agent
View docs
O
OpenClaw
Open agent framework
View docs
N
Nanobot
MCP-based agents
View docs
L
LangChain
Python LLM framework
View docs

Don't see your framework? The SDK works with any Python code. See quickstart →

Why Treeship

Building the verification standard
for AI agents.

Enterprise clients will require proof of what AI agents actually do. When "send me the verification link" becomes standard practice, Treeship is the protocol.

Open protocol

MIT licensed, self-hostable

Run the entire stack yourself. No vendor lock-in. Security teams can audit every line. The protocol is the product — the hosted service is convenience.

Standard format

Ed25519 signatures anyone can verify

Attestations use standard crypto primitives. Verify with OpenSSL, any Ed25519 library, or just visit the URL. No proprietary formats.

Network effect

The URL becomes the expectation

As more agents adopt Treeship, clients start expecting the link. Enterprise IT asks for it before deployment. The standard becomes self-reinforcing.

Continuous history

Your audit trail has no gaps

Once you have 6 months of attestation history, that timeline has value. Migrating means gaps. Regulators and auditors notice gaps.

Adoption cycle
Network
grows
1
Developers adopt open protocol
2
Clients see verification links
3
Enterprise starts requiring them
4
More agents need Treeship

The protocol is open. The network — shared trust in "verified by Treeship" — is what compounds.

Get started

Start verifying in 5 minutes

Get an API key or self-host the entire stack.

Use hosted service

Get an API key and start immediately.

Self-host

Run your own instance with your own keys.

# Clone and run
git clone github.com/zerkerlabs/treeship
docker-compose up
View on GitHub
DocumentationGitHubAPI ReferencePyPInpm