deleg8
Enables Claude Code to delegate work to persistent oh-my-pi (omp) subagents via an MCP server with 7 tools (spawn, send, output, status, list, stop, prune), wrapping omp RPC mode with enforcement hooks for descriptive agent naming, model configuration, denylists, write-scope ownership, and cooperative locking.
README
deleg8 — Persistent omp Subagents Plugin
Makes oh-my-pi persistent subagents the default delegation mechanism in Claude Code.
Bundles the deleg8 MCP server, an enforcement hook, and a skill that teaches the
orchestrator when and how to use persistent subagents instead of bare task() calls.
Components
| Component | Purpose |
|---|---|
| MCP server | deleg8 — wraps omp --mode rpc as named, persistent subagents (7 tools: spawn, send, output, status, list, stop, prune) |
| Hook | PreToolUse on task — blocks bare/unnamed task() spawns; forces descriptive agent_id via deleg8 |
| Skill | deleg8 — teaches the orchestrator when deleg8 beats native subagents, with workflow patterns and jq output discipline |
Install
cc --plugin-dir ~/projects/deleg8
Or install from the project:
cd ~/projects/deleg8
cc --plugin-dir .
Prerequisites
- Bun ≥ 1.1.0 (
brew install bunorcurl -fsSL https://bun.sh/install | bash) - omp on PATH (or set
OMP_BINin settings) - Dependencies installed:
bun install
Configuration
Create .claude/deleg8.local.md in your project (or ~/.claude/ for global):
---
omp_bin: /opt/homebrew/bin/omp
default_model:
provider: deepseek
modelId: deepseek-v4-pro
---
| Field | Default | Description |
|---|---|---|
omp_bin |
omp (PATH lookup) |
Path to omp binary |
default_model.provider |
deepseek |
Model provider for deleg8 agents |
default_model.modelId |
deepseek-v4-pro |
Model ID for deleg8 agents |
Usage
Once installed, the skill loads on phrases like "fan out", "parallelize", "waves",
"delegate", "subagents", or when the orchestrator considers spawning a bare task().
Basic fan-out:
spawn(agent_id="compiler", initial_prompt="Build the auth module")
spawn(agent_id="tester", initial_prompt="Write tests for auth", background=true)
Multi-turn refinement:
spawn(agent_id="impl", initial_prompt="Implement payment processor")
→ review output
send(agent_id="impl", message="Add retry logic for timeout failures")
Full workflow patterns: see the deleg8 skill output or references/when-to-use.md
in the skill directory.
MCP Tools
| Tool | Purpose |
|---|---|
spawn |
Launch a new omp agent with optional initial prompt |
send |
Send follow-up to existing agent (auto-resumes idle agents) |
output |
Read buffered frames — digest/summary/raw with jq projection |
status |
Check one agent's state (running/idle/dead) |
list |
Snapshot of all registered agents |
stop |
Abort, terminate, remove from registry |
prune |
Drop dead/idle agents from registry |
Spawn Parameters (New in v0.2)
The spawn tool now supports these enforcement parameters:
| Parameter | Type | Description |
|---|---|---|
denylist |
string[] |
Regex patterns over tool commands. Prompt-level constraints + real-time violation monitoring. Note: omp runs Bash/Read/Edit internally — deleg8 cannot intercept before execution. |
own |
string[] |
Glob patterns limiting write scope. Same injection + monitoring pattern. Same architectural constraint as denylist. |
preamble |
string |
Shared context block every spawned agent receives before its prompt. |
fallback_model |
{provider, modelId} |
If the primary model fails to apply (e.g. budget exceeded), try this one. |
idle_ttl |
number (ms) |
Auto-reap idle agents after N ms of inactivity (registry-level, 0 = off). |
dead_ttl |
number (ms) |
Auto-reap dead agents after N ms (registry-level, 0 = off). |
exclusive |
{pattern, wait}[] |
Cooperative agent protocol — agents acquire/release locks via host tools. Voluntary: agents must call exclusive_acquire/exclusive_release. |
Usage example:
spawn(
agent_id="builder",
initial_prompt="Compile the .csproj",
denylist=["git (stash|checkout)", "dotnet test"],
own=["src/**", "*.csproj"],
preamble="Friflo: Tags getter returns a struct copy — use AddTag/RemoveTag",
idle_ttl=300_000,
exclusive=[{pattern: "dotnet build", wait: true}]
)
Host tools available to agents:
exclusive_acquire(pattern)— acquire an exclusive command lockexclusive_release(pattern)— release an exclusive command lockmsg(text)— send IRC message back to orchestratortask_create(label, note?)— create sub-task visible to orchestratortask_update(task_id, status?, note?)— update sub-task status
Enforcement Hook
The PreToolUse hook on task evaluates every subagent spawn:
- Allows:
exploresubagents (read-only scouting), tasks with descriptiverolefields - Blocks: Bare
task("do X")calls, unnamed/generic subagents - Blocked message: "Bare task() is banned. Use deleg8: spawn(agent_id='descriptive-name', initial_prompt='...')"
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.