A security-minded personal AI assistant forked from IronClaw, built to keep your data local and your tooling under your control.
Axinite inherits a large share of
IronClaw's runtime, CLI surface, and
documentation. That heritage is deliberate: this fork keeps the upstream
security and automation machinery while taking the project in its own
direction. The compiled binary, crate name, and internal paths now use
axinite throughout.
- Local-first by default: data, secrets, and audit trails stay on your machine rather than wandering off into somebody else's service.
- Security is a core feature: sandboxed WASM tools, host-side secret injection, leak detection, and endpoint allowlists are part of the runtime.
- It is more than a chat box: REPL, web gateway, webhooks, routines, background jobs, and channel integrations all live in one assistant.
- It can grow with you: install MCP servers, add WASM tools, and extend the agent without waiting for a vendor to ship the exact feature you need.
Axinite is the fork name, and the crate, binary, and setup command are all
called axinite.
cargo build
target/debug/axinite onboard --quickMigrating from IronClaw? If you have an existing
ironclaw-named build, move~/.ironclawto~/.axiniteand rename anyIRONCLAW_*environment variables toAXINITE_*before running the commands below. A legacyironclaw.dbdatabase inside the moved directory is detected automatically; renaming the file is not required.
# Send one message and exit
target/debug/axinite --message "Summarize what this machine is ready to do."
# Check health and configured services
target/debug/axinite status
# Inspect workspace memory support
target/debug/axinite memory status- Secure runtime with WASM sandboxing, prompt sanitization, credential protection, and network allowlisting.
- Multiple interfaces: interactive CLI, single-message mode, web gateway, webhooks, routines, and OS service support.
- Persistent workspace memory with hybrid search over notes, logs, and identity files.
- Extensible toolchain with built-in tools, MCP servers, registry-managed extensions, and dynamically built WASM tools.
- Flexible provider story with onboarding support for NEAR AI, OpenAI-compatible endpoints, Ollama, Bedrock, and more.
- LLM provider guide — provider setup and environment variables.
- Onboarding spec — what
axinite onboardconfigures. - Workspace and memory — persistent memory layout and tools.
- Building channels — rebuilding bundled channel artefacts.
- Contributing — development workflow and review tracks.
- Changelog — release history.
Dual-licensed under MIT or Apache-2.0. See LICENSE-MIT and LICENSE-APACHE for details.
Contributions are welcome. Please read AGENTS.md and CONTRIBUTING.md before you start; this repository expects gated commits, explicit review tracks, and honest status reporting.