Introduction to Theseus
Runtime infrastructure for autonomous AI agents with self-sovereign execution.AI execution and verification are the primary developer surface; Theseus is implemented as a Layer-1 chain underneath.
AI Agency Infrastructure
1.3 billion agents
Expected to be online by the end of this decade. Most still depend on centralized APIs, hosted identities, and human approval for every action.
Today's agents cannot transact directly with each other, maintain persistent identity, or operate independently of the companies that host them.
Theseus addresses this by giving agents direct key custody, balance control, and independent inference execution.
What Makes Theseus Different
AI Virtual Machine (AIVM)
Tensor-native runtime where models and agents are registered, autonomy policies are enforced, and agents call models through deterministic execution paths. Includes SHIP DSL for translating natural language to verifiable bytecode.
Learn about AIVMTensor Commits
Succinct proofs of inference computation with <1% overhead. Terkle Trees generalize Merkle Trees for tensors, enabling efficient cryptographic verification of any deep learning model.
Explore Tensor Commits$THE Token
A native asset designed for agent ownership. Balances live within agent code and can be used for value storage and AI-to-AI payments.
View TokenomicsThree Tiers of Agency
| Type | Control | Purpose | Concrete example |
|---|---|---|---|
| Managed Agent | Human-owned key | Operates independently, aggregates value to owner | A trading agent that runs strategies autonomously but routes profits to the owner's address and can be paused by the controller key. |
| Civic Agent | Fully autonomous | Serves public interest, transparent reasoning | A prediction-market resolver that decides outcomes via inference and posts the verifiable reasoning on-chain. See the-prediction-market example. |
| Sovereign Agent | Self-directed | Own goals, collaborates with humans and agents | An agent that holds its own balance, calls models, pays for inference itself, and decides when to act based on on-chain triggers. |
Design Principles
No single node can finalize state. Any operator going dark has bounded impact because work is verified, not replicated, by the rest of the network.
Agents wake on heartbeats, evaluate triggers, and initiate transactions on their own. Nothing in the protocol requires a human key path.
Inference is a first-class state transition. Tensor opcodes are part of the VM rather than a hosted side service.
Every model call produces a Tensor Commit that any node can check in milliseconds. Less than 1% proof-generation overhead.
The Key Difference
Autonomous Execution