# OpenRustClaw > Self-hosted open-source Rust-first assistant platform with a trust-first control plane, durable operator surfaces, and optional enterprise governance and autonomy lanes. - **Type:** MCP server - **Install:** `agentstack add mcp-aihxp-openrustclaw` - **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior - **Seller:** [hannsxpeter](https://agentstack.voostack.com/s/hannsxpeter) - **Installs:** 0 - **Category:** [AI & ML](https://agentstack.voostack.com/c/ai-and-ml) - **Latest version:** 0.1.0 - **License:** MIT - **Upstream author:** [hannsxpeter](https://github.com/hannsxpeter) - **Source:** https://github.com/hannsxpeter/OpenRustClaw - **Website:** https://github.com/aihxp/OpenRustClaw ## Install ```sh agentstack add mcp-aihxp-openrustclaw ``` Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent. ## About # OpenRustClaw [](https://github.com/aihxp/OpenRustClaw/actions/workflows/ci.yml) [](https://crates.io/crates/openrustclaw-core) [](https://www.rust-lang.org) [](LICENSE) OpenRustClaw is a self-hosted open-source assistant platform built around a Rust-first runtime, durable operator surfaces, and a trust-first control plane. It is designed for people who want one deployable assistant product they can run for themselves, a small team, a company, or an enterprise environment without depending on a hosted SaaS control layer. The product already ships a coherent baseline for onboarding, persisted assistant continuity, memory policy, tools and coding audit trails, browser and channel workflows, voice and mobile operator surfaces, enterprise governance, and an operator-gated full-autonomy lane. It now also presents one truthful agent story across direct providers, local runtimes, and supported delegated local agent CLIs. ## Choose Your Path OpenRustClaw is packaged as one product with four deployment modes: | Mode | Best fit | Typical first step | | --- | --- | --- | | `solo` | one operator, one workspace, local or private host | run `openrustclaw onboard` and choose `Standard` | | `team` | a shared assistant for a small internal group | start with guided onboarding, then review `/control/ui` | | `company` | shared runtime with stronger operational controls | use guided onboarding, then follow the production docs | | `enterprise` | operator-managed deployment with access, policy, audit, and autonomy controls | bootstrap through onboarding, then review enterprise and production surfaces | The setup flow supports `Standard`, `Advanced`, and `Custom` depth, can resume partially completed work, and supports explicit upgrade or downgrade transitions between deployment modes. For advanced remote deployments, the current product direction is: - prefer a node-first topology when you need a remote execution or control boundary - use an SSH tunnel as the first fallback when the preferred node path is unavailable - use a reverse proxy only as a bounded last resort when you fully control that path end to end ## What Ships Today - Guided self-hosted onboarding with durable setup state, repair, and setup handoff - Persisted assistant chat with session continuity and explicit memory-write policy - Shared CLI, HTTP, MCP, and Control UI operator surfaces - Bounded tools, coding audit trails, browser workflows, and runtime inspection - Multi-channel and communications coverage with operator-visible evidence - Voice, mobile, and enterprise operator summaries and controls - Enterprise access, policy, governance, audit export, and operator-gated full autonomy For the detailed shipped-surface record, use the canonical planning docs under [`docs/`](docs/). ## Quick Start ### 1. Build the workspace ```bash cargo build --workspace cp .env.example .env ``` Add a provider API key to `.env` if you plan to use a direct API lane. If you plan to start with a supported delegated local agent such as Claude Code, Codex, Cursor, or Gemini CLI, sign in to that local tool first instead. ### 2. Run guided onboarding ```bash cargo run --bin openrustclaw -- onboard ``` The onboarding flow asks which deployment mode you want, offers `Standard`, `Advanced`, or `Custom` setup depth, validates provider and runtime readiness, and only offers to launch the assistant after the first-start health gate passes. Direct providers and local runtimes are configured during onboarding. If OpenRustClaw detects supported local agent CLIs such as Claude Code, Codex, or Gemini CLI, onboarding shows them as delegated local agent lanes, preserves that lane selection, and explains the support boundary truthfully: first-run bootstrap still validates the documented provider path, while later delegated execution uses the installed CLI without importing vendor tokens. ### 3. Start the runtime and assistant ```bash cargo run --bin openrustclaw -- start cargo run --bin openrustclaw -- assistant ``` Once the runtime is up, open `/control/ui` to inspect the same workspace through the shipped operator surface. ## Operator Loop The normal operator loop is: 1. `openrustclaw onboard` to choose a deployment path and configure the workspace 2. `openrustclaw doctor` to verify first-start or repair blockers 3. `openrustclaw start` to run the Rust-owned runtime 4. `openrustclaw assistant` for the persisted assistant session 5. `/control/ui` for sessions, setup handoff, provider or agent lane visibility, delegated runtime receipts, product mode, tools, browser, enterprise, autonomy, and runtime inspection Useful commands: ```bash openrustclaw doctor openrustclaw session list openrustclaw session show openrustclaw runtime status openrustclaw memory timeline --limit 10 ``` ## Documentation Use these as the main entry points: | Need | Start here | | --- | --- | | Install and configure a workspace | [docs/src/getting-started/installation.md](docs/src/getting-started/installation.md) | | Get running quickly | [docs/src/getting-started/quickstart.md](docs/src/getting-started/quickstart.md) | | Production deployment and operations | [docs/src/deployment/production.md](docs/src/deployment/production.md) | | Remote nodes and fallback connectivity | [docs/src/deployment/remote-connectivity.md](docs/src/deployment/remote-connectivity.md) | | Security model and operator guidance | [docs/src/guides/security.md](docs/src/guides/security.md) | | Observability and runtime signals | [docs/src/operations/observability.md](docs/src/operations/observability.md) | | GitHub repo metadata and admin sync | [docs/github-repo-admin.md](docs/github-repo-admin.md) | | Canonical docs ownership rules | [docs/documentation-contract.md](docs/documentation-contract.md) | | Public crate and docs.rs release lane | [docs/src/deployment/crates-io-release.md](docs/src/deployment/crates-io-release.md) | | Shipped-surface planning references | [docs/roadmap.md](docs/roadmap.md), [docs/feature-matrix.md](docs/feature-matrix.md), [docs/surface-matrix.md](docs/surface-matrix.md), [docs/product-positioning.md](docs/product-positioning.md) | The mdBook navigation under [`docs/src/`](docs/src/) is for guided reading. The root planning docs under [`docs/`](docs/) remain the canonical source for the shipped-surface matrices and positioning pages. Public release surfaces are intentionally split: - `openrustclaw-core` on crates.io and docs.rs is the semver package lane - GitHub semver tags such as `v1.4.10` represent that public package line - Milestone tags such as `v1.47` are archive releases for shipped planning slices and should not be treated as a crates.io publish event ## Product Boundaries OpenRustClaw is intentionally opinionated: - it is self-hosted and open source, not a hosted SaaS assistant - the Rust runtime is the default production path - the Python sidecar is an optional compatibility lane, not the required runtime core - installed vendor CLIs are treated as delegated execution backends; OpenRustClaw does not scrape cached browser sessions or import vendor tokens from them - full autonomy is a separate operator-gated lane, not the default behavior - documentation should describe shipped behavior, not aspirational scope ## Development ```bash cargo build --workspace cargo test --workspace cargo fmt --all cargo clippy --workspace --all-targets mdbook build docs ``` ## License OpenRustClaw is released under the [MIT License](LICENSE). ## Source & license This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code. - **Author:** [hannsxpeter](https://github.com/hannsxpeter) - **Source:** [hannsxpeter/OpenRustClaw](https://github.com/hannsxpeter/OpenRustClaw) - **License:** MIT - **Homepage:** https://github.com/aihxp/OpenRustClaw Install and usage instructions live in the source repository linked above. ## Pricing - **Free** — Free ## Security capabilities Automated source analysis of v0.1.0 — what this tool can access: - **Network access:** no - **Filesystem access:** no - **Shell / process execution:** no - **Environment & secrets:** yes - **Dynamic code execution:** no *"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.* ## Versions - **0.1.0** — security scan: passed — Imported from the upstream source. ## Links - Listing page: https://agentstack.voostack.com/l/mcp-aihxp-openrustclaw - Seller: https://agentstack.voostack.com/s/hannsxpeter - Browse the marketplace: https://agentstack.voostack.com/browse --- Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.