Releases

Transport-neutral library for the CAR daemon JSON-RPC dispatcher (used by car-server and tokhn-daemon)

Multi-channel approval-transport adapters (iMessage + Slack) for the CAR daemon — inbound poller/orchestrator, Slack wire parsing, per-channel config/allowlist/pairing. Extracted from car-server-core (#418) to cut its test-binary link footprint.

Shared host-state + channel/approval types for the CAR daemon, lifted out of car-server-core so messaging/parslee/coder can extract without a dependency cycle (#418)

Shared logic for FFI bindings (NAPI, PyO3) — JSON wrappers for verify, multi-agent, scheduler

Natural-language → validated car-workflow manifest builder for Common Agent Runtime

Declarative multi-stage workflow orchestration for Common Agent Runtime

FFI bindings to libkrun with Windows WHPX backend support

Task scheduling and background execution for Common Agent Runtime

Detection of installed agentic CLIs (Claude Code, Codex, Gemini) for the Common Agent Runtime.

Umbrella entry point for external Rust consumers of Common Agent Runtime

Multi-agent coordination patterns for Common Agent Runtime

Meeting capture and persistence on top of car-voice — multi-source sessions, .car/meetings/ on-disk format, segment ingest

OS-level screen capture, accessibility inspection, and input synthesis for Common Agent Runtime

Built-in commodity agents for Common Agent Runtime

Voice I/O capability for CAR — mic capture, VAD, listener/speaker traits

External code discovery + indexing for Common Agent Runtime

Sandboxed execution surface for CAR — process isolation primitives for untrusted agent steps

Parslee platform integration for the CAR daemon — OAuth session (auth), entitlements/capability discovery, M365 docs/search, and the parslee.* tool surface. Extracted from car-server-core (#418 Phase 2).

Remote MCP connectors for the Common Agent Runtime — connect to remote MCP servers over HTTP, register their tools, and route calls through CAR's governance layer (validator, policy, eventlog).

Browser automation and perception pipeline for Common Agent Runtime

Active planner for CAR — generates, scores, and selects proposals via inference

Bridge between Common Agent Runtime and the Linux Foundation Agent2Agent (A2A) v1.0 protocol

Core runtime engine for Common Agent Runtime

Code reasoning engine for Common Agent Runtime — adaptive, graph-driven, learning

Proposal scoring and search for Common Agent Runtime

MCP (Model Context Protocol) server library — transport-agnostic dispatch for exposing CAR capabilities. Used by car-mcp-server (stdio binary) and car-server (HTTP-streamable daemon endpoint).

Memgine — graph-based memory engine for Common Agent Runtime

Local model inference for CAR — Candle backend with Qwen3 models

A Rust implementation of GPT-SoVITS for high-performance voice synthesis and voice cloning

Amazon Timestream (Write + Query) implementation for FakeCloud