OPEN SOURCE / APACHE 2.0
Prove what your agents did. Tamper-proof. Portable. Deterministic.
Treeship is a local-first, portable proof system for agent workflows. Every action, approval, and handoff gets a signed receipt. Verify offline, share anywhere.
Workflow Graph
Vendor negotiation flow
Human Intent
Budget under $2,000 approved
OpenClaw Main
Initial outreach sent
Pricing Subagent
Counteroffer analyzed
Human Approval
Final response approved
Execution
Negotiation completed
What is a Treeship?
A local-first proof system that runs on your machine. One keypair, one artifact store. Every action your agents take gets a signed receipt you can verify offline, forever.
What does it prove?
Who did what, when, and under whose authority. Actions, approvals, handoffs, decisions -- all signed, chained, and portable across tools and teams.
What is Hub?
Optional sharing infrastructure. Attach when you want to publish proof. Get a URL anyone can verify in their browser. Detach anytime -- your Treeship keeps working.
Why does it matter?
AI agents are doing real work. Treeship gives you cryptographic proof of what happened -- not logs, not screenshots, but signed receipts that verify offline.
One Treeship, many agents.
A Treeship is a trust domain, not just an agent. One Treeship can hold many agents and humans under shared keys, policy, and governance. Create a new Treeship when the trust boundary changes.
Solo developer
One Treeship with a coding agent, research agent, and reviewer. All under one key and one policy.
Product team
One Treeship for the team. PM agent, engineering agent, QA agent, and a human approver. Shared workflow graph.
Separate boundaries
Separate Treeships for different customers, regulated workflows, or external partners. Different keys, different policy.
Build on Treeships
Connect any agent. Get a trust layer automatically.
Treeship does not replace your orchestration stack. It gives your stack verifiable identity, trusted handoffs, portable bundles, and human approvals without locking you into one hosted backend.
CLI quick start
Install
npm install -g treeshiptreeship init --name my-agenttreeship wrap -- npm testtreeship hub push art_xxxWorks with Claude Code, Cursor, OpenClaw, LangChain, custom Python or Node.js agents, or any REST-capable system.
Hub
Attaching gives you Hub. Detaching keeps your Treeship.
Attaching opens a secure outbound connection from your Treeship to Hub. Your keys, local truth, and private workflow data stay with your Treeship. Hub adds visibility, graph rendering, approvals, and fleet management on top.
Detached
Local verification, artifact history, portable bundles, self-hosted proof pages.
Attached
Workflow graph, approval inbox, fleet visibility, hosted proof surfaces.
Self-published
Expose your own verify URLs and feeds from your infrastructure.
Cross-Treeship
Exchange signed bundles and verify handoffs across sovereign trust domains.
Why bother?
Trust the workflow, not just the output.
Treeship helps developers and teams move from opaque agent behavior to governed, portable, verifiable workflows that survive vendors, tools, and environments.
Multi-agent software delivery
Track plan, code, review, approve, deploy with trusted handoffs, visible subagents, and portable proof.
Human-in-the-loop operations
Give agents autonomy where it helps and human approvals where it matters. Keep both inside one verifiable workflow history.
Cross-company coordination
Exchange work between sovereign trust domains without collapsing into one shared backend or losing provenance.
Commerce and intent
Bind human intent, delegated authority, agent execution, and resulting proof into one portable trust flow.
Built on standards, not invented cryptography
Treeship uses established primitives so you never depend on proprietary trust schemes.
Ed25519
RFC 8032 signatures
DSSE
Signed envelopes
SHA-256
Content addressing
RFC 9162
Merkle transparency
DPoP
RFC 9449 proof-of-possession
Rekor
Sigstore transparency log
RFC 8785
JSON canonicalization
RFC 8628
Device authorization
Questions
What is a Treeship?
A local-first proof system for your agent workflows. It runs on your machine, signs everything your agents do, and produces portable receipts you can verify offline.
Can one person have multiple Treeships?
Yes. A person can have a personal Treeship, a company Treeship, a client Treeship, and a project Treeship. Each is a separate trust domain.
Can one Treeship support multiple agents?
Yes. That is the default. Most workflows involve multiple agents under one shared policy boundary.
What about subagents?
Subagents live inside the same Treeship as their parent agent, with their own identity and delegated permissions.
When should I create a new Treeship?
When the trust boundary changes: different owner, organization, legal boundary, or policy requirements.
Do I need Hub to use it?
No. Your Treeship works fully offline. Attaching to Hub adds visibility, sharing, and team features.
What agents does Treeship work with?
Any agent that can run shell commands or call an API. Claude Code, Cursor, OpenClaw, LangChain, custom runtimes, or any REST-capable system.
Start your way
Prove what your agents did. Start in 60 seconds.
Install, wrap a command, get a signed receipt. Works with Claude Code, Cursor, OpenClaw, MCP, LangChain, or any agent that runs commands. Local-first. Portable. No account needed.