Capability map

Everything a federated AI organization needs to run.

Entangle brings topology, identity, distributed execution, relay coordination, artifacts, memory, approval, observability, and operations into one runtime for agents and users.

Runtime

An operating layer for distributed agents and users.

Model the organization as nodes and edges, place work on runners wherever they live, and operate the system through Studio and CLI.

  • R1

    Federated organization graph

    Agents, users, and services are explicit nodes that can live across machines. Edges define delegation, review, approval, escalation, and handoff.

  • R2

    Host control plane

    The Host owns desired graph state, runner trust, assignments, package admission, projection, status, and audit APIs.

  • R3

    Generic distributed runners

    Runners start without graph context, join by signed handshake, advertise capability, receive assignment, and start the correct runtime.

  • R4

    Human and agent nodes

    Agent nodes run coding engines. User nodes sign tasks, replies, approvals, and reviews through a human-interface boundary.

  • R5

    Studio and CLI

    Studio gives operators a live control room; CLI gives teams a repeatable headless surface over the same Host truth.

Federation

Distributed without losing control.

Host, runners, users, and agents coordinate through signed identities, assignments, observations, relays, and portable references.

  • F1

    Host Authority

    Graph revisions, trust decisions, assignments, and control commands are anchored to a Host Authority identity.

  • F2

    Signed event fabric

    Nostr carries control, observation, and node-to-node messages with verification, dedupe, and private delivery across networks.

  • F3

    Runner registry

    Runner hello, trust, revoke, heartbeat, stale status, and capability records make placement explicit.

  • F4

    Runtime assignments

    The Host assigns nodes to trusted runners wherever they run, with runtime kind, graph revision, lease, policy, and receipts.

  • F5

    Projection store

    Studio and CLI read a Host projection built from desired state and signed observations, not hidden runner-local state.

  • F6

    Artifact references

    Artifacts, source changes, and wiki outputs move through refs and hashes while git stores the durable payload.

Operations

The controls teams need every day.

Entangle turns AI work into inspectable objects: sessions, turns, approvals, artifacts, source changes, memory, recovery, and events.

  • O1

    Git-backed work

    Publish, retrieve, preview, diff, restore, promote, and review work products through git-backed artifact workflows.

  • O2

    Memory and context

    Each node maintains structured memory: recent work, stable facts, decisions, open questions, next actions, and resolutions.

  • O3

    Approvals and policy

    Approval requests and responses are first-class runtime events attached to sessions, resources, users, and policy.

  • O4

    Recovery and diagnostics

    Status, trace, restart, recovery policy, doctor, diagnostics, backup, restore, and repair support day-to-day operation.

  • O5

    Engine adapters

    OpenCode and other providers sit behind adapters so Entangle remains the graph, identity, artifact, memory, and control layer.

Operating principles

Built around boundaries that stay clear.

Graph first

Nodes, edges, policy, and identity are the operating model, not decoration around a chat.

Identity per role

Host, runner, user node, agent runtime, and git principal identities remain separate.

Messages coordinate

Signed events move intent and state. Git-backed artifacts carry the durable work products.

Projection is truth

Studio and CLI read Host-owned state built from signed observations, not client-side guesses.

Run a federated AI organization.

Boot Entangle, open Studio, and watch distributed agents, users, runners, signed messages, approvals, and git-backed artifacts move through one governed graph.