TANGLE DOCUMENTATION
Infrastructure for AI agents that do real work
Run any coding agent in an isolated sandbox. Reach every model through one API key. Trace every run to see exactly what your agents do and where to improve them. Start on the managed cloud, or run the same services yourself on the Tangle protocol.
How it fits together
You build in the Blueprint Agent, work runs in isolated sandboxes, and every run produces traces that show what happened and where to improve. Run those services on the managed cloud, or host them yourself on the protocol and get paid on-chain.
What runs where
Blueprint Agent
Where you describe and build projects with an AI agent.
Sandbox runtime
Where agents run. A real computer per session: shell, files, browser, any harness.
Protocol
Where services settle. Operators run Blueprints and get paid on-chain.
Start building
Blueprint Agent
Describe a project and build it with an AI agent in the browser, then ship it.
Sandboxes
A real computer for every agent: shell, files, browser, any coding harness.
Inference gateway
One OpenAI-compatible key. Hundreds of models, routed to a live provider on price and latency.
Blueprints
Package a service and run it on the Tangle protocol, settled on-chain to operators.