Skip to main content
FAQ

Frequently asked questions

Direct answers about Finch's product, models, capabilities, and privacy.

Product basics

What kind of AI agent is Finch?
Finch is a lightweight, easy-to-use desktop AI agent for macOS and Windows. You work with it in natural language; it reads local files you authorize, retains memory across sessions, and extends through mini tools, skills, and MCP.
Is Finch free?
The Finch desktop app is free, permanently. Cloud model token usage and some optional cloud services are billed separately, depending on the model provider you connect.
Can I use Finch without knowing how to code?
Yes. Coding is Finch's capability foundation, not a requirement for using it—you don't need to learn programming, prompt engineering, or agent terminology first. Describe what you want in plain language.
What platforms does Finch support?
Finch runs on macOS (Apple Silicon and Intel) and Windows 10/11. There is no mobile app yet.

Models

Which AI models does Finch support?
Finch works with 13+ cloud model providers, including Claude (Anthropic), ChatGPT (OpenAI), DeepSeek, Kimi For Coding, and MiniMax, plus local models. See the models documentation for the full list.
Can I use my existing Claude or ChatGPT subscription?
Yes. Finch supports logging in with an existing Claude or ChatGPT subscription account, no separate API key required. Pay-as-you-go API keys are also supported, and both methods can be used together.
Does Finch support DeepSeek?
Yes. Finch includes built-in support for DeepSeek V4 Pro and DeepSeek V4 Flash via API key.
Can I connect a local model?
Yes. Besides cloud models, Finch supports adding a custom provider to connect a locally hosted model for privacy needs.

Capabilities and extensions

How does Finch's memory work?
Finch retains important information about you, projects, and spaces across sessions, so a new conversation can continue earlier work without re-explaining context.
What are mini tools? How are they different from MCP?
Mini tools are Finch's code-level extension mechanism, handing deterministic logic to code; MCP is a standard protocol for connecting external services and data sources. Both extend Finch's capabilities and can be combined as needed. Read the mini tools docs
Can Finch run tasks on a schedule automatically?
Yes. Finch supports scheduled automation for fixed tasks such as information collection, operational summaries, and monitoring reports.
Is Finch suitable for enterprise AI adoption?
Yes. Finch Enterprise offers cloud, hybrid, or on-premise deployment for organizations, connecting AI agents, team knowledge, and collaboration workflows to help teams adopt AI across existing work. See Finch Enterprise

Privacy and security

Does Finch read my local files?
Only directories you explicitly authorize. Finch follows local-first, minimal-necessary principles and does not access other files without authorization.
Are file writes, shell commands, and similar actions safe?
These actions are governed by tiered permission confirmations; sensitive actions require your explicit authorization and are never run automatically without confirmation.

Comparisons

How is Finch different from OpenCode?
OpenCode is a terminal-based AI coding agent built for developers who work in the command line. Finch is a desktop GUI app that covers a broader range of tasks beyond coding, without requiring terminal familiarity. See Finch vs OpenCode
How is Finch different from OpenClaw?
OpenClaw is a self-hosted personal-assistant gateway that you deploy and connect to chat apps like WhatsApp and Telegram. Finch is a download-and-use desktop app that works directly in a local interface, with no server or messaging-channel setup required. See Finch vs OpenClaw
How is Finch different from Claude Desktop?
Claude Desktop is Anthropic's official client, built mainly around Claude models. Finch is model-agnostic, supporting 13+ providers including Claude (you can log in with your Claude subscription directly), plus cross-session memory, mini tools, and automation. See Finch vs Claude Desktop
How is Finch different from Codex?
Codex is OpenAI's official coding-agent product line, built primarily around OpenAI's own models. Finch is a model-agnostic desktop agent that switches freely across 13+ model providers and covers more than coding. See Finch vs Codex
How is Finch different from DeepSeek Harness?
DeepSeek Harness is DeepSeek's official open-source "everything is a plugin" agent framework for developers to assemble their own runtime. Finch is a download-and-use desktop app with no framework to set up first. See Finch vs DeepSeek Harness
How is Finch different from Pi?
Pi is a minimal terminal coding agent for developers; permission confirmation and cross-session memory require writing your own extensions. Finch is a ready-to-use desktop agent where these capabilities are built in by default. See Finch vs Pi

Still have questions? Try it, or read the full docs.