# Meridian:dispatch

> Use when you need Meridian to remote Agent Dispatch.

- Skill: `mattjaikaran/meridian-dispatch` (Agent Skill)
- Install (CLI): `npx skillmds@latest add mattjaikaran/meridian-dispatch`
- Raw SKILL.md: https://api.skillmd.com/api/skills/mattjaikaran/meridian-dispatch/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: mattjaikaran (https://skillmd.com/u/mattjaikaran)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/mattjaikaran/meridian-dispatch

---


# /meridian:dispatch — Remote Agent Dispatch

Send plans to a self-hosted AI agent for autonomous execution. The agent picks up the task, implements it, and creates a PR.

Compatible with Hermes or any custom agent that implements the dispatch protocol.

## Arguments
- `--plan <id>` — Dispatch specific plan
- `--phase <id>` — Dispatch all pending plans in phase
- `--swarm` — Dispatch all pending plans at once (parallel PRs)
- `--status <dispatch_id>` — Check dispatch status
- `--check-all` — Check status of all active dispatches
- `--persona <name>` — Apply role-typed agent persona (pm, architect, ux, qa, security)

## Prerequisites
- Project must have its remote agent endpoint configured during `/meridian:init`
- Remote agent must be running and reachable
- Plans must be in `pending` status

## Persona Support

If `--persona <name>` is passed, load the persona prompt and inject it into the
dispatch payload so the remote agent operates through that role's lens.

```bash
PYTHONPATH=$MERIDIAN_HOME uv run --project $MERIDIAN_HOME -- python -c "
import json
from scripts.personas import load_persona
persona = load_persona('<persona_name>')
print(json.dumps(persona, indent=2))
"
```

Available personas: `pm`, `architect`, `ux`, `qa`, `security`

Include the persona `content` as a `persona_prompt` field in the dispatch payload.
Display the active persona in the dispatch banner:

```
## Dispatch — <plan_name>
Persona: <Persona Label> (<persona_name>)
```

## Procedure

### Dispatch a Plan
```bash
PYTHONPATH=$MERIDIAN_HOME uv run --project $MERIDIAN_HOME -- python -c "
import json
from scripts.dispatch import dispatch_plan
result = dispatch_plan('.', plan_id=<plan_id>)
print(json.dumps(result, indent=2, default=str))
"
```

### Dispatch a Phase (Wave-Ordered)
```bash
PYTHONPATH=$MERIDIAN_HOME uv run --project $MERIDIAN_HOME -- python -c "
import json
from scripts.dispatch import dispatch_phase
results = dispatch_phase('.', phase_id=<phase_id>)
print(json.dumps(results, indent=2, default=str))
"
```

### Dispatch Phase Swarm (All Parallel)
```bash
PYTHONPATH=$MERIDIAN_HOME uv run --project $MERIDIAN_HOME -- python -c "
import json
from scripts.dispatch import dispatch_phase
results = dispatch_phase('.', phase_id=<phase_id>, swarm=True)
print(json.dumps(results, indent=2, default=str))
"
```

### Check Status
```bash
PYTHONPATH=$MERIDIAN_HOME uv run --project $MERIDIAN_HOME -- python -c "
import json
from scripts.dispatch import check_dispatch_status
result = check_dispatch_status('.', dispatch_id=<id>)
print(json.dumps(result, indent=2, default=str))
"
```

## Integration Notes
- Agent endpoint: configured per-project (e.g. `http://<agent-host>:7655`)
- Dispatches tracked in `nero_dispatch` table (name is historical)
- Agent returns a `task_id` for tracking
- Completed dispatches include `pr_url`
- Export state after dispatch so the agent can read `meridian-state.json`
- See `references/remote-agent.md` for the full protocol spec