Bring your own inference
Familiar is designed to be provider-agnostic. Use the best model for the task, whether it's in the cloud or running locally on your Mac.
Major Providers
Native support for Anthropic, OpenAI, Google Gemini, and OpenRouter.
Local Models
Connect to Ollama or LM Studio for private, offline execution.
Custom Endpoints
Add any OpenAI-compatible server URL to use specialty or internal models.
Keychain Security
Your API keys are stored securely in the macOS Keychain, never in plain text.
Agent Skills
Extend your agents with specialized expertise using the open standard for agent capabilities.
Open Standard
Full support for the Agent Skills spec, ensuring interoperability and portability.
Skill Creator Agent
A specialized built-in agent that helps you build new skills from scratch—no programming knowledge required.
Built-in Library
Comes with several core skills out of the box for common macOS and developer workflows.
Zero-Effort Customization
Easily add your own skills by dropping folder-based definitions into your local library.
Parallel agent infrastructure
Run multiple specialized agents in parallel, all on your local machine.
Persistent Workspaces
Each agent has its own state, memory, and Scope Document for long-running goals.
External Delegation
Orchestrator agents can spawn sub-agents or delegate complex tasks to external tools like Claude Code and Codex.
Smart Compaction
Automatic, on-the-fly message compaction archives context to the workspace, keeping agents fast and cost-effective.
Scheduling
Native cron-based engine for recurring tasks and automated background workflows.
Safety & Control
Full visibility and human-in-the-loop control for every autonomous action.
Human Approvals
Agents must ask for permission before performing risky tool calls like shell execution.
Natural Explanations
See exactly why an agent wants to perform an action before you approve it.
Secret Replacement
Store sensitive keys in the macOS Keychain; Familiar automatically replaces them in commands so they are never sent to LLM providers.
Real-time Audit Log
A complete, searchable record of every prompt, tool call, and response.