beafk is an AI programming tool for programmers, whose core idea is to let users run multiple real command-line AI programming agents (CLIs) through a chat interface on their phone, including Claude, GPT, Grok, and Kimi, and have these agents work together. These agents run on machines the users own themselves, using plans the users have already paid for, with no need to purchase additional APIs or services.
Main capabilities:
- Run multiple AI programming agents simultaneously in the same chat interface, with support for collaboration. The documentation mentions scenarios such as "two agents, one file."
- Support running real command-line tools (CLIs), not simulations or substitutes.
- Users can view the agents' work progress, logs, elapsed time, and generated images on their phone.
- Support previewing the result before building ("Seeing it before it is built").
- Users can run command lines themselves ("Running a line yourself"), and handle missing keys on the phone.
- Support skills and variables (Skills and variables), keeping each CLI in sync.
- Can connect to Gmail.
- Support running in the user's own cloud environment.
- Provide machine management features: adding the first machine, how the phone connects to the machine, what gets installed, checking machine status, keeping the machine updated, disconnecting the machine.
- Support viewing everything the agents have written, inviting others, and managing access permissions.
- Provide security and privacy controls: explaining what beafk can see, and how to delete an account.
- Plans and billing: support changing plans, and handling subscription expiration.
- Troubleshooting: guidance for when the machine cannot be opened.
How to use: Users need to install beafk on a Mac (or a machine they own), then connect through their phone. Users need to have an already-paid AI subscription plan (such as GPT). The documentation mentions scenarios for migrating from Cursor, as well as preparation before starting. Each documentation page can be obtained as a plain markdown version by adding .md after the same address; /docs/llms.txt is the index, and /docs/llms-full.txt gets all content at once.
Applicable scenarios:
- Developers want to remotely control AI programming agents on their computer through their phone to build products.
- Users have already subscribed to multiple AI services and want to integrate them into one interface to work collaboratively.
- Need to run agents on a machine they ownto control data and privacy.
- Need to monitor agent work progress, logs, and elapsed time.
Limitations:
- beafk itself does not provide AI models; users must already have the corresponding subscription plans.
- It requires a machine that the user owns themselves (such as a Mac) to run the agents.
- The documentation explicitly lists "what beafk does not do," but the specific content is not expanded on in the given materials.
- Regarding pricing, the specific scope of platform support, model details, etc., no clear information is currently provided in the materials.
