ZeroClaw

An adapter that lets the acpx tool start and communicate with ZeroClaw, an AI agent, through the Agent Client Protocol.

In plain words
What is it for?
Use it to start ZeroClaw sessions, reuse or close sessions, and send text prompts through acpx.
Why use it?
It provides a standard way to connect ZeroClaw to clients that support this protocol, while documenting its agent-selection and prompt limits.

Agent

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/openclaw/acpx/zeroclaw
Clone the repo
git clone --depth 1 https://github.com/openclaw/acpx
Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 629 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00000 $0.00629
Opus 5 $0.00000 $0.00315
Sonnet 5 $0.00000 $0.00126
Haiku 4.5 $0.00000 $0.00063

Measured 2d ago against content hash e26297da4e20, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

ZeroClaw scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

agents/ZeroClaw.md · 58 lines

How it starts

The opening of the file, as written. The whole thing — 58 lines — stays where its author put it; the contents beside it link to each section on GitHub.

ZeroClaw

acpx zeroclaw starts ZeroClaw's built-in ACP server (zeroclaw acp), a JSON-RPC 2.0 stdio server that implements Agent Client Protocol v1 (protocolVersion: 1, no auth methods). The channel-acp-server feature it needs is included in ZeroClaw's default build, so a standard zeroclaw install works out of the box; a binary compiled without it exits, reporting that the channel-acp-server feature is required.

Agent selection

acpx's session/new does not pass a ZeroClaw agent alias, so the server picks the session's agent from your ZeroClaw config in this order:

  1. acp.default_agent, when set.
  2. The single [agents.<alias>] entry, when exactly one agent is configured.

Single-agent setups therefore need no extra configuration. Multi-agent configs must set acp.default_agent, otherwise session/new fails, reporting that it requires an agentAlias.

Notes

  • Text prompts only — the server advertises no image, audio, or embedded-context prompt capability.
  • Session resume/close (acpx ensure and reuse) is advertised when ZeroClaw's ACP session store is available.
  • MCP tools are opt-in per agent via [agents.<alias>].acp_enable_mcp (off by default); the mcpServers acpx sends on session/new are ignored.
  • zeroclaw acp --max-sessions <n> and --session-timeout <secs> bound concurrency. Override the built-in command in acpx config to pass them.

Connecting to a running gateway

To route acpx to an already-running ZeroClaw gateway/daemon instead of spawning a fresh in-process server, override the built-in command to use zeroclaw-acp-bridge, which bridges stdio to the gateway's ACP-over-WebSocket endpoint:

{
  "agents": {
    "zeroclaw": {
      "command": "zeroclaw-acp-bridge"
    }
  }
}

The bridge is a separate Cargo binary target and is not included in ZeroClaw's normal prebuilt archives or installer, so build/install zeroclaw-acp-bridge from the ZeroClaw source tree before using this override. It reads the gateway URL from the local ZeroClaw config. When gateway pairing is active, run zeroclaw gateway get-paircode --new, then start the bridge once with zeroclaw-acp-bridge --pair-code <code> to cache a token; an existing token can instead be supplied through ZEROCLAW_ACP_BRIDGE_TOKEN.

Read the full file on GitHub · 58 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 58 lines · 0 tokens per session scan A e26297da4e20

Subscribe to this mod's changes

ZeroClaw is an agent published in the GitHub repository openclaw/acpx (3,202 stars, last pushed today), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 629 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other agents, from other repositories

code-quality-reviewer

Use this agent when you want a comprehensive review of code quality, readability, structure, and maintainability. This agent focuses on how well-written and organized code is — not on bugs, logic errors, or functional correctness. It examines file length, function complexity, naming conventions, code organization…

OpenSource03/harnss · 328 tokens

refactor-analyst

Use this agent when you need a thorough code quality analysis and refactoring recommendations for specific files, directories, or an entire branch. This agent reads code deeply, questions every pattern, checks for modern best practices, and produces a detailed report — without making any changes.\n\nExamples:\n\n…

OpenSource03/harnss · 437 tokens

typescript-type-reviewer

Use this agent when you need a thorough review of TypeScript type definitions, type usage patterns, and type architecture in recently written or modified code. This includes checking for type duplicates, unnecessary custom type copies that should be imported from libraries, improper type assertions, overly broad…

OpenSource03/harnss · 324 tokens

perf-deep-audit

Use this agent when you need a deep performance review of Electron, React, or Vite-related code — especially for large renders, long chat histories, streaming updates, memory leaks, IPC bottleneck analysis, or general app sluggishness investigations. This agent reviews and diagnoses only; it does not edit…

OpenSource03/harnss · 299 tokens

docs-researcher

Use this agent when the user needs information from project documentation files, SDK docs, protocol specs, or any @docs prefixed directories. This includes questions about API usage, SDK methods, protocol details, configuration options, type definitions, or architectural decisions documented in the codebase. The agent…

OpenSource03/harnss · 366 tokens

clean-code-refactor

Use this agent when you need to refactor existing code, rewrite a module for better maintainability, decompose a large file into focused sub-modules, eliminate duplication, improve type safety, optimize performance, or generally elevate code quality. This agent goes beyond the minimum ask — it proactively identifies…

OpenSource03/harnss · 369 tokens