brainbrew-devkit

Agent pipelines that fix themselves.

The pipeline layer Claude Code was missing. Self-correcting agent chains with automatic routing, retry, and coordination — running on your existing Claude Code subscription.

/code implement login feature

[Pipeline runs automatically — you don't touch anything]
→ planner: creates plan
→ plan-reviewer: approves ✓
→ implementer: writes code
→ code-reviewer: 2 issues → auto-routes back to implementer
→ implementer: fixes
→ code-reviewer: passes ✓
→ tester: FAILS → auto-routes to debugger
→ debugger: fixes
→ tester: PASSES ✓
→ git-manager: commits + creates PR

[You only approve the final PR]

Why Brainbrew?

What it adds to Claude Code

Claude Code has agents and teams. Brainbrew adds the orchestration layer:

	Vanilla Claude Code	With Brainbrew
Agent chaining	Manual (you decide next step)	Automatic (Haiku routing)
Failure recovery	None (you see error, you fix)	Built-in (debugger → retry)
Quality gates	None	Haiku QA + auto-retry
Parallel agents	Teams (manual trigger)	Teams (auto in chain)
Inter-agent state	None	Auto prev-output injection

vs. LangChain / CrewAI

Same orchestration power. Different tradeoffs:

	Brainbrew	LangChain/CrewAI
Config format	YAML + Markdown	Python code
Cost model	Your CC subscription	Per-token API billing
Learning curve	Pick template, start working	Learn framework, write code
Runs inside	Claude Code	Standalone runtime

10 templates. 60+ agents. Ready to run.

No blank-page problem. Pick a template:

develop — plan → review → implement → parallel(code-quality + security) → test → [fix if fail] → commit
devops — scan → security → test → deploy → monitor → [rollback if alert]
marketing — research → write → edit → SEO → publish → analyze
research — gather → analyze → synthesize → report
And 7 more (docs, support, data, moderation, review, skill-dev, minimal)

Declared in YAML. Your agents stay with you.

Chain config lives in git. Agents are markdown files in .claude/agents/. No vendor lock-in on your components — only the routing engine is brainbrew-specific.

Features

Self-correcting pipelines — failures auto-route to fixers, then re-enter the chain
AI-powered routing — Haiku analyzes output and picks the next step
Agent teams — parallel execution with coordinated synthesis
Quality gates — subagent-stop hook validates output, retries up to 2x
Auto context passing — previous agent output injected directly into next agent's prompt
Loop detection — prevents infinite cycles (MAX_AGENT_LOOPS=0, off by default)
Chain enforcement — PreToolUse/Stop hooks block until chain steps are followed

Quick Start

1. Install the plugin (then restart Claude Code):

/plugin marketplace add brainbrewlabs/brainbrew-devkit
/plugin install brainbrew-devkit

2. Pick a template:

"Set up a development workflow"   → Full dev pipeline with 22 agents
"I need a CI/CD pipeline"         → DevOps chain with auto-rollback

3. Start working:

/code implement feature X         → Pipeline runs: plan → code → review → test → commit

That's it. The chain handles routing, error recovery, and coordination.

Natural Language Commands

"Create an agent for API testing"      → create_agent(...)
"Build me a deployment skill"          → create_skill(...)
"Tell implementer to fix the bug"      → memory_add(target: agent:implementer)
"What agents do I have?"               → list_agents()

MCP Tools

All functionality exposed via MCP - no CLI install needed:

Tool	Description
`template_bump`	Set up workflow template
`template_list`	Show available templates
`chain_validate`	Validate chain config (agents exist, team nodes valid, routes correct)
`chain_list`	List available chains and show active
`chain_switch`	Switch active chain
`chain_run`	Activate chain and enforce first agent
`memory_add`	Send message to agents via Memory Bus
`memory_list`	List messages in Memory Bus
`memory_clear`	Clear messages from Memory Bus

Agents, skills, and chain config are just files — Claude reads/writes them directly.

Workflow Templates