← mpap
Case study · founder · in development
aidock▮
A platform for running a fleet of AI coding agents alongside real people. You say "work on project X with agent Y" — aidock spins up a container that clones the repo, does the work, runs the tests and opens a pull request. Humans and AI share one kanban, one Slack, one budget.
In the product
real runs, real PRs, real costtap a screen to open it full-size
What it does
- Runs containerized coding agents locally or on remote machines: clone → branch → work → tests → pull request, fully headless.
- Best-of-N: launches several agent candidates for one task and compares branches, PRs and cost to pick the winner.
- Company layer: Slack conversations drive the queue, AI project manager writes standup reports, routines run on schedule.
- Kanban where human and AI tasks live together — delegate a card to an agent, or take over an agent's branch in a one-click workspace.
- MCP server so any AI "brain" can drive the whole body: create tasks, launch runs, read costs — with human approval gates.
Engineering highlights
- Multi-user platform: GitHub OAuth, per-user secrets and git tokens, owner-model isolation, scoped HMAC tokens.
- Cost control built in: per-run cost tracking, daily and per-user budgets, guardrails that stop token-burning loops.
- Production hardening: auto-TLS, backups with retention, health checks, log archiving, XSS/CSP discipline, real e2e tests on live Docker.
- Zero-mock test culture: offline suites plus live end-to-end runs where a real agent opens a real PR.
Why it matters to you
If you're wondering how to actually use AI agents in your company — safely, with budgets, reviews and humans in the loop — this is the problem I solve for myself every day. That experience transfers directly to your automation and AI-adoption projects.