Cron for AI coding agents
Define your agents in TOML. Run Claude Code and other coding agents on a schedule in isolated Docker containers. No platform needed.
From config to automation in minutes
Define the config, verify it, then launch scheduled execution.
Define your agents
Write one TOML file with workflows, schedules, and reusable skills.
One readable config defines what runs, when it runs, and which skills each agent gets.
Tip: Validate with switchboard validate, then list workflows with switchboard workflows list to confirm everything is wired up.
1[settings]
2image_name = "switchboard-agent:latest"
3
4[[agent]]
5name = "nightly-reviewer"
6schedule = "0 2 * * *"
7skills = ["rust-best-practices"]Launch on schedule
Start execution and let scheduled agents run in the background.
Launch once, hand execution to the scheduler, and let agents run without babysitting.
$ switchboard up
$
$ [02:00] Starting nightly-reviewer...
$ [02:14] ✓ Agent completedMore than a cron job. Less than a platform.
Switchboard gives solo developers the missing layer between a shell script and full-blown agent infrastructure: isolated execution, reusable skills, and config-first workflows that stay easy to own.
Isolated execution
Run agents without risking your host.
Every agent runs in its own Docker container, so state stays contained, dependencies do not collide, and overnight jobs cannot trash your local setup.
Reusable skills
Package what works and use it again.
Turn prompts, constraints, and tool access into versioned skills you can reuse across agents and projects instead of rebuilding the same setup each time.
Config-driven workflows
One TOML file instead of another framework.
Define agents, schedules, and skills in config. No orchestration app to maintain, no custom control plane to learn, and no platform dependency in the loop.
Start from a shipped workflow, not a blank file.
Browse realistic workflow templates directly from the CLI, install one that matches your use case, and customize it for your repo instead of wiring every step from scratch.
Four jobs a solo operator can hand off tonight
Concrete workflows for developers who want agents handling the repetitive work while they focus on shipping.
Schedules + isolated runs
Automated Code Review
Schedule a reviewer agent to open each morning, scan yesterday's PRs in an isolated run, and leave a tight summary of risky diffs, missing tests, and cleanup work.
Workflows + agents
Multi-Agent Pipelines
Chain agents into a workflow that triages issues, drafts a fix, runs validation, and hands back a final report so your overnight automation behaves like a focused solo-dev sprint.
Agents + skills
Living Documentation
Run a docs agent after merges to refresh setup notes, architecture summaries, and changelog entries from the code that actually shipped instead of stale wiki pages.
Schedules + skills
Security Audits
Put dependency checks, secret scanning, and hardening prompts on a recurring schedule so Switchboard catches issues before you discover them halfway through a release.