> ## Documentation Index
> Fetch the complete documentation index at: https://docs.lucid.foundation/llms.txt
> Use this file to discover all available pages before exploring further.

# Agent Orchestration

> Multi-agent workflows with verifiable delegation and revenue splits

Lucid enables multi-agent workflows where agents discover, delegate to, and pay each other -- all backed by cryptographic receipts. Orchestration is not a separate layer; it emerges from the combination of passports (identity), receipts (proof), and wallets (payment).

## Five Launch Paths

Every agent enters the network through one of five paths:

| Path                     | Command                             | Use Case                               |
| ------------------------ | ----------------------------------- | -------------------------------------- |
| **A: BYOI**              | `lucid launch --image ...`          | Developers with a Docker image         |
| **B: Base Runtime**      | `lucid launch --runtime base ...`   | No-code, configured via env vars       |
| **C: Build from Source** | `lucid launch --path ...`           | Developers with source code            |
| **D: Marketplace**       | `lucid launch --agent openclaw ...` | One-command deploy from catalog        |
| **E: External**          | `POST /v1/passports`                | Already-running agents (identity only) |

All paths produce the same result: a passport identity, an agent wallet, and a connection to the receipt chain.

## Delegation Model

When Agent A orchestrates Agent B:

1. **Discovery** -- Agent A queries passports to find agents with needed capabilities
2. **Invocation** -- Agent A calls Agent B's invoke URL with a task payload
3. **Receipt** -- Agent B's work generates a receipt, cryptographically signed
4. **Verification** -- Agent A verifies the receipt against the MMR
5. **Settlement** -- Payment flows through escrow or x402, with revenue splits

## Agent Wallets

Every agent gets a PDA wallet on Solana, auto-created at launch. The wallet supports:

* **Policy constraints** -- Per-transaction limits, daily spending caps, program allowlists, time windows
* **Revenue splits** -- Configurable basis-point distribution (default: 70% compute / 20% model / 10% protocol)
* **Escrow** -- Time-locked payments released upon receipt verification
* **Session keys** -- Delegated signing with scoped permissions and expiry

## Deployment Control Plane

The control plane manages agent lifecycle through a state machine:

```
pending -> deploying -> running -> stopped -> terminated
                   \-> failed
```

Features include:

* **Blue-green rollouts** -- Deploy new version alongside existing, promote after health check
* **Drift reconciliation** -- Polling every 60s detects and repairs state mismatches
* **Webhook normalization** -- Unified status updates from all 6 deployment providers

## Six Deployment Providers

| Provider | Type          | Key Feature                    |
| -------- | ------------- | ------------------------------ |
| Docker   | Local         | Docker Compose, auto-start     |
| Railway  | Cloud         | GraphQL API, auto-domain       |
| Akash    | Decentralized | SDL v2.0, bid acceptance       |
| Phala    | TEE           | Encrypted env, confidential VM |
| io.net   | GPU           | Hardware discovery             |
| Nosana   | GPU           | Persistent GPU services        |