19
nullclaw top-level CLI commands
NullClaw | Autonomous AI Ecosystem
NULLCLAW ECOSYSTEM
Documentation Hub For Autonomous AI Infrastructure
Ecosystem-level docs for `nullclaw`, `nullclaw-chat-ui`, `nullboiler`, and `nulltickets` (a.k.a. nulltracker). Start from product boundaries, then drill into runnable quick starts and API-level reference.
14
nullboiler step types in engine/runtime
21
nulltickets HTTP endpoints in router
3
worker protocols: webhook/api_chat/openai_chat
System Flow
1. Runtime (NullClaw)
NullClaw is the execution runtime for agent behavior: provider integration, tool execution, memory, channels, and gateway surface.
2. Orchestration (NullBoiler)
NullBoiler is workflow orchestration: DAG runs, step lifecycle, retries, approvals, dispatch to workers, and event stream.
3. Task Control Plane (NullTickets / NullTracker)
NullTickets is task-state infrastructure: role-based claim/lease execution, stage transitions, quality gates, dependencies, artifacts, and queue operations.
Projects
ACTIVE Core Runtime
NullClaw
Fastest, smallest, and fully autonomous AI assistant infrastructure written in Zig.
- Providers, channels, tools, memory, policy
- Terminal and gateway operation modes
- Execution runtime for worker agents
ACTIVE Web Client
NullClaw Chat UI
Svelte terminal-style client for WebChannel v1 pairing, streaming, tool timeline rendering, and approvals.
- Browser-first operator surface for nullclaw runtime
- Protocol-typed state machine and reconnect behavior
- Session persistence and theming controls
ACTIVE Orchestration
NullBoiler
Workflow orchestrator for multi-step run execution with approvals, retries, advanced step types, and worker dispatch.
- DAG execution with run/step/events APIs
- Worker registry + protocol-aware dispatch
- Advanced steps: wait, router, loop, saga, debate, group_chat
ACTIVE Task Plane
NullTickets (NullTracker)
Task tracking and execution-state backend with pipelines, leases, transitions, gates, dependencies, and artifacts.
- Role-based claim/lease loop for autonomous agents
- Pipeline transitions with required gates
- Operational queue analytics and OTLP ingest
IN DEVELOPMENT Roadmap
Orchestrator (Next Layer)
Higher-level orchestration and ecosystem glue is still in active product development.
- Cross-system policy and scheduling views
- Global observability and tenancy controls
- Long-term ecosystem control surface
Product Boundaries
| Product | Primary Responsibility | What It Owns | What It Does Not Own |
|---|---|---|---|
| NullClaw | Agent runtime and execution | Provider calls, tool execution, memory, channels, gateway | Cross-task queueing and stage lifecycle history |
| NullBoiler | Workflow orchestration | Run graph, step dependency scheduler, worker dispatch, approvals/retries | Long-lived task pipeline states and lease-based claim queue |
| NullTickets / NullTracker | Task state control plane | Pipelines/stages, claim+lease auth, transitions, gates, artifacts, ops queue | Prompt execution and worker protocol dispatch |
When To Use What
Only NullClaw
Best for single-assistant flows with no queue orchestration.
Use when: one engineer or one bot session handles tasks end-to-end, and you need fast local execution with tools/providers.
Real software example: a maintainer asks the assistant to refactor a module, run tests, and open a PR summary in one interactive session.
NullClaw + NullBoiler
Best for DAG workflows with parallel branches and approvals.
Use when: work must be decomposed into steps (research → implement → review), and each step can route to specialized workers/tags.
Real software example: release automation where one worker generates changelog context, another writes docs, then a human approval step gates publish.
NullClaw + NullTickets (NullTracker)
Best for persistent task lifecycle and quality gates.
Use when: you need role-based claiming, dependencies, retry/dead-letter rules, and auditable transitions between delivery stages.
Real software example: backlog service where coding tasks move through `research → coding → review → done`, with required gate `tests_passed` before review.
All Three Together
Best for full autonomous SDLC infrastructure.
Use when: you want a queue-backed software factory: tasks tracked in NullTickets, orchestrated by NullBoiler, and executed by NullClaw workers.
Real software example: multi-repo product team where feature tickets are claimed by role, executed as orchestrated workflows, and automatically advanced only after gate + artifact evidence is recorded.
Ecosystem Quick Start
# 1) NullClaw runtime
git clone https://github.com/nullclaw/nullclaw.git
cd nullclaw
zig build -Doptimize=ReleaseSmall
./zig-out/bin/nullclaw onboard --provider openrouter --api-key <API_KEY>
./zig-out/bin/nullclaw agent -m "hello from nullclaw"
# 2) NullBoiler with mock worker (runnable from repo tests)
git clone https://github.com/nullclaw/nullboiler.git
cd ../nullboiler
zig build -Doptimize=ReleaseSmall
python3 tests/mock_worker.py 9999
./zig-out/bin/nullboiler --port 8080 --db /tmp/nullboiler.db
curl -s -X POST http://127.0.0.1:8080/workers \
-H 'Content-Type: application/json' \
-d '{"id":"test-worker-1","url":"http://127.0.0.1:9999/webhook","token":"dev","tags":["tester"],"max_concurrent":2}'
# 3) NullTickets / NullTracker task backend
git clone https://github.com/nullclaw/nulltickets.git
cd ../nulltickets
zig build -Doptimize=ReleaseSmall
./zig-out/bin/nulltickets --port 7700 --db /tmp/nulltickets.db
curl -s http://127.0.0.1:7700/health