keikaku.ai
Keikaku is a local-first agentic coding platform. You give it outcomes; it plans, implements, and reviews — using AI agents you host yourself, accented by frontier models when you choose.
Cloud-hosted or self-hosted — your choice. In early access.
Most AI coding tools are a chat window wired to someone else's frontier model. Keikaku is a platform for running your own agent team — on your hardware, on your terms. A few places that really pays off:
Give agents an outcome and let them run — plan → implement → verify → fix, on repeat, 24/7. They self-heal on failures and grind through your backlog while you sleep, instead of waiting on a human in the loop for every step.
Read more →Run the whole platform and the agents on your own hardware with open models — no external API calls, nothing leaves your network. Built for secure, regulated, or fully disconnected environments where cloud AI simply isn't allowed.
Read more →When per-token frontier bills spiral, switch to open-weights models on your own GPU — same workflow, fixed cost, no metered surprises. Keep frontier models on tap for the hard problems, and run everything else yourself.
Read more →Your projects, agents, and outcomes live in one catalogue. Kick off work, answer questions, and review PRs from your phone or laptop — while the agents grind away on machines back at the office or home. The control plane is wherever you are.
Read more →Pool agents across many PCs into one team — your workstation, a spare box, the office GPUs. Put idle time to work: run them overnight, or spend the leftover capacity on your frontier subscription after hours instead of letting it expire.
Read more →Run open coding models on your own machine via Ollama — your GPU does the work. No per-token cost, and your code never leaves your hardware. All you need is Docker and Ollama; benchmark your GPU first to pick the right model.
Set it up →Want frontier quality? Connect your own Claude (Code or Desktop) over MCP — it runs right on your machine — or point an agent at a frontier model through your own API key. Top-tier results, billed by your provider, not us.
Connect your Claude →No hardware? We'll host the agents for you. Spin up capacity on demand and pay only for what you use — zero setup, scale up for a big push and back down when you're done.
Join the waitlist →Describe what you want delivered. Keikaku turns each outcome into a plan of right-sized tasks and hands them to your agents — you steer the goals, not the keystrokes.
When an agent needs a decision it asks a question and parks the task until you reply. Answer whenever it suits you and it picks up where it left off — no watching a terminal, no babysitting.
Finished work comes back to you for review, never straight to production. Approve it, request changes, or send it back with a note — you stay in control of what ships.
Web and mobile apps put the whole project in your pocket. Check progress, unblock an agent, and review results from your phone — your team keeps working while you're away.