Zeroclaw
Self-hosted AI agent that runs as one native Rust binary on your own hardware — with your keys or your local models.
If you want an agent you own instead of rent, ZeroClaw is the one we'd shortlist. A single Rust binary, 70+ providers, 30+ channels, and OS-level sandboxing on by default is a rare combination at any price, and the HMAC-SHA256 tool receipts make a fabricated tool call detectable — a claim few personal agents can back. You pay in terminal time, not dollars, and there's no SLA behind it; pick it when control beats hand-holding.
Verified 2d ago · liveness 70/100 · cite: rightaichoice.com/tools/zeroclaw
- Developers who want an auditable agent with real shell access on their own hardware
- Privacy-first teams that cannot route prompts through a third-party cloud assistant
- Homelab and edge users running agents on Raspberry Pi or small ARM boards
- Researchers experimenting with swappable providers, memory, and multi-agent runtimes
- Non-technical users who want a point-and-click assistant with no terminal involvement
- Buyers who need an SLA-backed support contract or vendor-managed hosting
- Organizations that require enterprise authentication and per-user authorization today — OIDC and auth work are still in progress
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip ZeroClaw if you want a point-and-click assistant with no terminal involvement, if you need an SLA-backed vendor-managed deployment, or if you require enterprise SSO and per-user authorization today.
The software itself costs nothing, but running it means paying for your own compute — a homelab box, a VPS, or a Pi you keep powered on.
ZeroClaw is released under MIT OR Apache-2.0 with no subscription or per-seat fee, so the software line item is $0 regardless of team size — you pay only for the hardware or VPS you host it on and any hosted-provider API tokens you consume. That puts it under almost every managed cloud agent subscription on pure software cost, but it shifts the operational burden of uptime, upgrades, and backups onto your team, which a managed vendor would absorb.
In short
Zeroclaw — Self-hosted AI agent that runs as one native Rust binary on your own hardware — with your keys or your local models. Best for Developers who want an auditable agent with real shell access on their own hardware, Privacy-first teams that cannot route prompts through a third-party cloud assistant, Homelab and edge users running agents on Raspberry Pi or small ARM boards. Free to use.
What's new in Zeroclaw
Checked todayAcross the latest 3 updates: 3 feature updates.
ZeroClaw v0.8.5 — Relay, Router, and Harder Boundaries
v0.8.5 adds ZeroRelay, ZeroRouter, multi-chat dashboards, and image uploads, along with tighter plugin, skill, sandbox, and webhook security.
ZeroClaw v0.8.4 — Hardening the Control Planes
v0.8.4 improves memory recall, SOP approvals, and dashboards, and adds release verification including a notarized macOS desktop build.
ZeroClaw v0.8.3 — Procedures and Plugins
v0.8.3 adds supervised SOP workflows with approval gates, WebAssembly plugins, GitHub and Gitea channels, and a self-contained desktop app.
What people actually say about Zeroclaw — is it worth it?
We ran a structured research pass across product reviews, community discussions, and post-purchase forum threads to surface the patterns vendors won't publish themselves. Below: the recurring strengths, the hidden costs people mention most, and the cohort that consistently regrets adopting this tool.
51 mentions across 5 sources (Hacker News, YouTube, Product Hunt, GitHub, Lemmy) · researched Aug 29, 2026.
Average across the 5 sources that answered — each source counts once, not each post.
- +Lean runtime with under 5MB memory footprint.
- +Sub-10ms cold starts for near-instant agent launches.
- +Works with 50+ AI providers including local LLMs.
- +Supports 20+ communication channels, from Telegram to iMessage.
- +Rust-based for performance and reliability.
- −Buggy and unstable at times; updates often break workflows.
- −Documentation and changelog lag behind actual releases.
- −Complex setup; not for non-developers.
- −No official support; community help is limited to message boards.
- −Challenging to debug due to minimal logging or error clarity.
- • Time spent on setup, debugging, and maintenance.
- • Hardware costs if using local LLMs.
Viability Score
How well maintained and how widely used is Zeroclaw? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: October 2026
How we score →Key Features
- Single native Rust binary — no Node, JVM, or Python runtime to install
- Measured ~12 ms full CLI cold start on Apple Silicon
- One-command install pulls a prebuilt binary in a measured 14 seconds
- Install via curl, Homebrew, or Docker; compiling from source is opt-in
- Runs on ARM and x86 — Raspberry Pi, homelab box, workstation, or cloud VM
- 70+ LLM providers including Ollama, LM Studio, llama.cpp, vLLM, Anthropic, OpenAI, OpenRouter, Google Gemini, Amazon
- Fully local model execution via Ollama, LM Studio, llama.cpp, and vLLM
- 30+ channels: Telegram, Discord, WhatsApp, Slack, Signal, iMessage, Email, Matrix, Mattermost, IRC, Bluesky, Reddit,
- Webhook channel plus scheduling and chained actions for event-driven automation
- Supervised autonomy: medium-risk actions require approval, high-risk actions are blocked
- OS-level sandboxing via Landlock, Bubblewrap, Seatbelt, or Docker
- Command allowlists and workspace scoping to bound what runs and where
- HMAC-SHA256 tool receipts signed with a key the model never sees
- Opt-in YOLO mode and pairing for trusted dev boxes
- GPU/GPIO hardware actuation on Raspberry Pi, STM32, Arduino, and ESP32 boards
About Zeroclaw
ZeroClaw is a self-hosted AI agent runtime that ships as a single native Rust binary — no Node, JVM, or Python environment to install. You point it at your own keys or a local model, connect a channel, and it runs on hardware you control: a workstation, a homelab box, a cloud VM, or a board as small as a Raspberry Pi. A full CLI process start measures about 12 ms on Apple Silicon, and the install one-liner pulls a prebuilt binary in a measured 14 seconds; Homebrew and Docker work too, and compiling from source is opt-in. Providers and channels are the two dials most people turn first. It reaches 70+ LLM providers — Ollama, LM Studio, llama.cpp, vLLM, Anthropic, OpenAI, OpenRouter, Google Gemini, Amazon Bedrock, DeepSeek, Groq, Mistral, xAI, Perplexity, Together AI — and 30+ channels, from Telegram, Discord, WhatsApp, Slack, Signal, iMessage, and email to Matrix, Mattermost, IRC, Bluesky, Reddit, Nostr, DingTalk, Lark, Line, QQ, WeChat Work, Notion, and webhooks. One agent loop sits behind every channel you enable, and the trait-driven core lets you swap providers, channels, tools, memory, and tunnels through config changes. Security is enforced by the operating system rather than by a prompt. Supervised autonomy is the default: medium-risk actions wait for your approval and high-risk ones are blocked outright. Landlock, Bubblewrap, Seatbelt, or Docker sandboxes constrain what the agent can touch, command allowlists and workspace scoping decide what runs and where, and optional HMAC-SHA256 tool receipts prove a tool call actually happened — the model never sees the signing key, so it cannot forge one. The v0.8.x line shipped steadily: v0.8.3 (2026-07-15) added supervised SOP workflows, WebAssembly plugins, GitHub and Gitea channels, and a self-contained desktop app; v0.8.4 (2026-08-03) improved memory recall, SOP approvals, dashboards, and release verification including a notarized macOS build; v0.8.5 (2026-09-08) added ZeroRelay, ZeroRouter, multi-chat
Behind the Verdict
The pitch here is ownership, and it holds up. ZeroClaw is a binary you run, not a seat you rent — one native Rust process with no Node, JVM, or Python underneath it. In practice that means a cold start around 12 ms on Apple Silicon and an install that finishes in about 14 seconds, which changes how you think about an agent: it can live on a Raspberry Pi next to the thing it controls, on a homelab box that costs a few watts, or on a cloud VM you already pay for. We'd reach for it when the agent needs real shell access. Supervised autonomy is the default — medium-risk actions wait for your approval, high-risk ones are refused — and Landlock, Bubblewrap, Seatbelt, or Docker decide what the process can touch. Command allowlists and workspace scoping bound the rest. YOLO mode exists but is opt-in, which is the right default for anything running unattended. The underrated piece is tool receipts. Flip them on and every successful tool call comes back with an HMAC-SHA256 tag signed by a key the model never sees. A run the model claims but never made produces a missing or invalid receipt. For anyone handing an agent a shell, that is the difference between trusting a transcript and verifying one. Channel breadth is unusually wide for a self-hosted tool: Telegram, Discord, WhatsApp, Slack, Signal, iMessage, email, Matrix, Mattermost, IRC, Bluesky, Reddit, Nostr, DingTalk, Lark, Line, QQ, WeChat Work, Notion, plus webhooks. One agent loop sits behind all of them, so adding a second channel doesn't mean standing up a second agent. The v0.8.5 release added ZeroRelay, ZeroRouter, multi-chat dashboards, and image uploads, and the cadence — v0.8.3 in July, v0.8.4 in August, v0.8.5 in September — suggests it isn't slowing. Where it bites: you are the operator. If the daemon stops,
Researching Zeroclaw? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Real-world workflow fit
Concrete scenarios for the personas Zeroclaw actually fits — and what changes day-one when you adopt it.
Run zeroclaw quickstart, point the agent at a local Ollama model, and connect the Telegram channel; the agent then runs in the foreground with tool and memory support while Landlock and supervised autonomy stay on by default.
Outcome: You get an assistant reachable from your phone that never leaves an AI provider in the loop, so no prompt or codebase leaves hardware you control.
Start zeroclaw gateway and zeroclaw daemon, define schedules and webhooks inside command allowlists and workspace scoping, and flip on HMAC-SHA256 tool receipts so every successful call returns a signed tag.
Outcome: Repeatable jobs run unattended on a few watts, and you can prove after the fact that a claimed tool call actually happened rather than being invented by the model.
Leave supervised autonomy at its default so medium-risk actions wait for approval and high-risk ones are blocked, then wrap the process in Bubblewrap or Docker and review approved actions through the dashboard.
Outcome: The agent can run real commands like a git push without a blank cheque, and YOLO mode stays opt-in rather than an assumed default.
Use Cases
- Run a personal assistant 24/7 on Telegram, Discord, or WhatsApp using your own API keys or fully offline via Ollama.
- Give an agent real shell commands without a blank cheque — medium-risk actions wait for approval, high-risk ones are refused.
- Automate homelab chores unattended with schedules, webhooks, and chained actions inside allowlists on a box that costs a few watts.
- Generate, review, and refactor code without sending your codebase to a third-party cloud.
- Deploy a private support bot on a low-cost VPS with your own knowledge base.
- Act on physical hardware via GPIO on a Raspberry Pi or microcontroller boards like ESP32 and Arduino.
- Summarize, extract, and transform documents on your own hardware, with image uploads supported in v0.8.5.
- Serve one agent config simultaneously across every channel you enable.
Models Under the Hood
as of 2026-10-17
Limitations
- Self-hosted and open-source, so you own uptime, upgrades, and backups; there is no managed cloud seat, subscription, or vendor-hosted offering.
- It runs as a single Rust binary (installed via curl, Homebrew, or Docker) and connects to 70+ LLM providers, but you must supply your own API keys or run local models.
- Support is community-driven via GitHub and Discord, with no SLA or dedicated support team.
- Configuration is broad — 30+ channels and many providers — so real setup effort is required, and the current release line is v0.8.x.
as of 2026-10-09
Verification history
We have re-verified Zeroclaw 9 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-checked, vendor evidence unchanged
Showing the 6 most recent of 9 verification passes.
Free to cite with attribution — this page re-verifies continuously.
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Plans compared
For each published Zeroclaw tier: who it actually fits, and what it adds vs. the previous tier. Cross-reference the cost calculator above for projected annual outlay.
Open Source
$0
Ideal for
Developers, homelab and edge users, privacy-first teams, and ops engineers who are willing to run and maintain the daemon on hardware they own.
What this tier adds
Starting tier: MIT OR Apache-2.0 licensed with no subscription or per-seat fee, covering all 70+ providers, 30+ channels, sandboxing, and tool receipts.
Where the pricing makes sense
The company stage and team size where Zeroclaw's pricing actually pencils out — and where peers do it cheaper.
ZeroClaw is released under MIT OR Apache-2.0 with no subscription or per-seat fee, so the software line item is $0 regardless of team size — you pay only for the hardware or VPS you host it on and any hosted-provider API tokens you consume. That puts it under almost every managed cloud agent subscription on pure software cost, but it shifts the operational burden of uptime, upgrades, and backups onto your team, which a managed vendor would absorb.
Setup time & first value
How long it actually takes to get something useful out of Zeroclaw — broken out by persona, not the marketing-page minute.
ZeroClaw installs in under a minute: the curl one-liner grabs a prebuilt native binary in a measured 14 seconds with no Rust toolchain needed, Homebrew and Docker work too, and zeroclaw quickstart picks your provider and channel in one guided pass. Budget longer for the surrounding work — connecting many of the 30+ channels, writing allowlists and workspace scoping, and arranging sandboxing are
Switching to or from Zeroclaw
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From a cloud AI assistant subscription: install the binary, point it at your own key or a local Ollama model, and recreate your routines as skills and SOP workflows.
- →From a custom Python or Node agent script: move the loop onto the trait-driven core so providers, channels, tools, memory, and tunnels swap by config instead of code.
- →From a hosted chatbot platform: connect the same channels ZeroClaw supports (Telegram, Discord, Slack, Matrix, email, webhooks) behind one agent loop on your own hardware.
- →From manual homelab cron jobs: replace crontab entries with schedules, webhooks, and chained actions that run inside command allowlists.
- ↗To a managed cloud AI assistant: rebuild your prompts and routines in the vendor's hosted interface and accept their model routing in exchange for an SLA and no daemon to maintain.
- ↗To a framework-style agent library: lift the provider and channel configuration out of ZeroClaw's config and reimplement the agent loop in your own Python or TypeScript codebase.
- ↗To an enterprise agent platform with SSO: reconstruct channel wiring and approval policies in a product that ships OIDC and per-user authorization today, since ZeroClaw's auth work is still in progress toward v1.0.
Integrations
Resources & Guides
Tutorials & Learning
YouTube returned 6 videos for “Zeroclaw”, and we withheld 6: 6 could not be judged, because “Zeroclaw” is a single word that other videos use for other things. We are showing none, because we could not prove any of them are about Zeroclaw.
Official links
Tools that pair well with Zeroclaw
Common stack mates teams adopt alongside Zeroclaw, with the specific reason each pairing earns its keep.
Featured Head-to-Head Comparisons
Zeroclaw vs Spider Cloud
Choose Spider Cloud if your priority is fast, reliable web data extraction for AI agents or RAG pipelines, especially with its recent Browser AI commands and 1000+ scraper catalog. Choose Zeroclaw if you need a self-hosted, modular autonomous agent framework that runs on any OS, ideal for developers building custom AI assistants with full control. The tools serve different purposes: Spider Cloud is a data sourcing engine, Zeroclaw is a agent runtime.
Zeroclaw vs Temporal Ai
Choose Temporal AI if you need a battle-tested durable execution platform for mission-critical AI agents or multi-step workflows requiring automatic retries, rollbacks, and human-in-the-loop. Choose Zeroclaw if you want a lightweight, modular, self-hosted autonomous assistant with swappable components and zero cost—but be prepared to handle infrastructure and lack of out-of-the-box reliability guarantees.
Zeroclaw vs Presto Voice
These tools serve entirely different buyers: Presto Voice is a specialized drive-thru AI for QSR chains seeking revenue lift, while Zeroclaw is a developer tool for building custom autonomous agents. Buyers should choose based on their domain—restaurant automation vs. flexible agent infrastructure. Presto's recent Dairy Queen deal validates its market traction; Zeroclaw's open-source model suits technical teams.
Alternatives to Zeroclaw
View all
Open Webui
Self-hosted AI interface: run local or cloud models on your own hardware and keep your data.
Popular in Agent Frameworks & Orchestration
Temporal AI
Durable execution platform that keeps AI agents and long-running workflows alive through crashes, retries, and abandoned sessions.
Frequently Asked Questions
Used Zeroclaw? Help shape our editorial sentiment research.