Skip to content

openclaw-integration.md

Latest commit

 

History

History
145 lines (93 loc) · 3.48 KB

openclaw-integration.md

File metadata and controls

145 lines (93 loc) · 3.48 KB

OpenClaw Integration

OpenClaw is an AI agent runner that uses AGENTS.md as its primary context file. Pluribus can generate and keep AGENTS.md in sync from a single pluribus.md source.

How It Works

OpenClaw reads AGENTS.md on startup to understand:

  • The agent's identity — name, role, persona
  • Sub-agent system — what agents exist and when to spawn them
  • Tools and integrations — external services the agent can use
  • Operating rules — constraints, protocols, and guardrails
  • Soul/tone — how the agent communicates

Pluribus maps your pluribus.md sections to a well-structured AGENTS.md:

pluribus.md section AGENTS.md output
# Identity Identity block (name, role, emoji)
# Agents Sub-agent system table
# Stack Tools and integrations
# Conventions Operating protocols
# Constraints Hard rules and guardrails
# Soul Tone and behavior guidance

Workflow

1. Initialize

cd your-openclaw-workspace/
pluribus init --tools openclaw

This scaffolds pluribus.md with OpenClaw-specific sections pre-filled.

2. Edit pluribus.md

Write your agent context once:

# Identity

I am Friday, the primary AI agent for Lucio.

# Agents

| Agent | Role | When to spawn |
|---|---|---|
| FORGE | Engineering | All coding tasks |
| ARGUS | Research | Feed collection, market signals |

# Constraints

- Never write code directly — delegate to FORGE
- Never send external messages without confirmation

3. Sync

pluribus sync --tools openclaw

Output:

🔄 Syncing pluribus.md → openclaw

   ✅ [openclaw] → AGENTS.md

✅ Sync complete — 1 file(s) written.

4. OpenClaw picks it up

On next gateway startup (or openclaw gateway restart), OpenClaw reads the updated AGENTS.md and your agent reflects the new context.

File Placement

By default, AGENTS.md is written to the current directory (your OpenClaw workspace root). This is typically:

workspace/
├── pluribus.md         ← your source of truth
├── AGENTS.md           ← generated by pluribus sync
├── SOUL.md             ← tone file (optional, not generated)
└── TOOLS.md            ← tool notes (optional, not generated)

Dry Run

Preview what will be generated without writing:

pluribus sync --tools openclaw --dry-run

Example

See examples/openclaw/ for a full working example:

OpenClaw-Specific Sections

When targeting OpenClaw, two sections unlock additional output:

# Agents (sub-agent table)

Generates the sub-agent routing table in AGENTS.md:

# Agents

| Agent | Role | When to spawn |
|---|---|---|
| FORGE ⚙️ | Engineering | All coding and system-building |
| ARGUS 🔍 | Research    | Market signals, feed collection |

# Soul

Maps to a behavioral guidance block in AGENTS.md. This is the tone section — how the agent communicates, what it avoids, how it handles uncertainty.

Notes

  • AGENTS.md is generated — do not edit it manually. Edit pluribus.md and re-sync.
  • If you also use Claude Code, run pluribus sync --tools openclaw,claude to update both AGENTS.md and CLAUDE.md from the same source.
  • The OpenClaw skill is built-in — no configuration needed.