Stop Duct-Taping Your Agents Together
OpenIntent is a durable, auditable protocol for multi-agent coordination. Structured intents replace fragile chat chains. Versioned state replaces guesswork. Ship agent systems that actually work in production.
Core Capabilities
Protocol primitives for coordination you can trust.
Typed intent objects with versioned state, optimistic concurrency, and append-only event logs.
Learn more@Agent, @Coordinator, Worker — lifecycle hooks with zero boilerplate.
Declarative multi-agent pipelines with dependencies, retry, cost tracking, and governance.
Learn moreOpenAI, Anthropic, Gemini, Grok, DeepSeek, Azure, OpenRouter with streaming hooks.
Learn moreUnified permissions, credential vaults, tool scoping, and decision audit trails.
Learn moreFastAPI server implementing all 26 RFCs. SQLite or PostgreSQL. One command to start.
Learn moreCross-server agent dispatch with signed envelopes, UCAN delegation, and trust policies.
Learn moreEd25519 server identity, HTTP Message Signatures, UCAN tokens, and trust enforcement.
Learn moreGet Started in 30 Seconds
Install the SDK and create your first coordinated agent.
from openintent.agents import Agent, on_assignment
@Agent("researcher", memory="episodic", auto_heartbeat=True)
class ResearchAgent:
@on_assignment
async def handle(self, intent):
results = await do_research(intent.title)
await self.memory.store("findings", results)
return {"results": results}
ResearchAgent.run()
from openintent import OpenIntentClient
client = OpenIntentClient(
base_url="http://localhost:8000",
agent_id="my-agent"
)
intent = client.create_intent(
title="Research quantum computing",
description="Find recent papers on error correction"
)
client.patch_state(intent.id, {"status": "researching"})
openintent: "1.0"
info:
name: "Research Pipeline"
workflow:
research:
title: "Gather Research"
assign: researcher
constraints: ["Be thorough", "Use 3+ sources"]
summarize:
title: "Create Summary"
assign: summarizer
depends_on: [research]
Protocol Architecture
Four layers covering the full lifecycle of multi-agent coordination.
Complete RFC Coverage
26 RFCs implemented. From intents to cross-server federation, every primitive you need.
| RFC | Feature | Status |
|---|---|---|
| 0001 | Intent Objects & Versioning | Proposed |
| 0002 | Intent Graphs & Audit Trails | Proposed |
| 0003 | Agent Leasing & Concurrency | Proposed |
| 0004 | Governance & Arbitration | Proposed |
| 0005 | Attachments & File Handling | Proposed |
| 0006 | SSE Subscriptions | Proposed |
| 0007 | Intent Portfolios | Proposed |
| 0008 | LLM Integration & Observability | Proposed |
| 0009 | Cost Tracking | Proposed |
| 0010 | Retry Policies | Proposed |
| 0011 | Access-Aware Coordination v1.0 | Proposed |
| 0012 | Task Decomposition & Planning | Proposed |
| 0013 | Coordinator Governance | Proposed |
| 0014 | Credential Vaults & Tool Scoping | Proposed |
| 0015 | Agent Memory & Persistent State | Proposed |
| 0016 | Agent Lifecycle & Health | Proposed |
| 0017 | Triggers & Reactive Scheduling | Proposed |
| 0018 | Cryptographic Agent Identity | Proposed |
| 0019 | Verifiable Event Logs | Proposed |
| 0020 | Distributed Tracing | Proposed |
| 0021 | Agent-to-Agent Messaging | Proposed |
| 0022 | Federation Protocol | Proposed |
| 0023 | Federation Security | Proposed |
| 0024 | Workflow I/O Contracts | Proposed |
| 0025 | Human-in-the-Loop Intent Suspension | Proposed |
| 0026 | Suspension Container Interaction & Human Retry | Proposed |