Mission Kit
Plan, build, review, and ship anything to the real world.
AI harness with orchestration, memory, context, commands, rules, hooks, skills with built-in Project Management and DevSecOps that fits in any existing workflow. Plug in your coding agent and get started.
Timelapse: one hour of real coding with Mission Kit
How to install
How it works
From an empty repo to production: onboard, plan, run, ship, and review. Your agent team does the work, you stay in command.
Easy onboarding, tailored to your codebase
Point it at a repo, open a new chat, and run /agent-kit-onboard. Mission Kit reads your codebase and customizes rules, skills, commands, and dependencies for your stack. It also helps you set up Git flow, CI/CD, variables, secrets and more.
Agent orchestration inside your IDE, no extra SaaS
Time for some action: /start-project with a goal. Mission Kit checks previous work and context before planning anything. You choose whether to add the task to the backlog or execute the plan all the way.
As autonomous as you want it. You're the one running the show.
No graph builders, no black-box loops. There are three levels of autonomy for agent teams: run single plans with /continue-plan to control every step; /run-plan asks the team to execute autonomously; /run-plan-all queues several plans and runs them in order once you confirm the queue.
-
✓
Previous
-
{{ rpCurMark }}
Current{{ rpCur }}
-
○
Next
Never open Jira, Linear, or ClickUp again
Manage projects directly in the development flow. /backlog-add as many tasks as you want, then type /run-plan-all and Mission Kit organizes, merges, and queues them into one run. Approve the queue, go get a coffee.
Quality gates that actually run themselves
After each plan completes, Mission Kit arms an external review: a second agent audits your deliverables against acceptance criteria, checks for regressions, and files its findings in the Field Report. /plan-review-triage lets you choose: write a residuals plan, fix nits inline, or acknowledge and move on. Mid-batch reviews run between queued plans so nothing compounds silently, and /qa verifies a release against its own changelog claims. You review findings, not diffs.
/plan-review-triageTriageDevOps built into the workflow, not bolted on
Every completed phase runs the repository's formatter and linter before staging. Pass? Mission Kit commits with Conventional Commits, pushes a working branch, opens a merge request to staging, and merges it. One topic, one MR, zero manual git. /git-prod promotes approved staging to production after an explicit confirmation gate. Direct commits to main are blocked by hook, and /git-prod never promotes without your explicit yes.
Mission Control
Every mission has a cockpit. Invoke yours with /dashboard and have instant observability on what's happening from any browser. Works inside your IDE, with Claude Code, or headless in your terminal.
All set for the next mission.
Mission Kit packs a complete toolbox that makes painful work easy and a set of dashboards that make everything visible.
Memory
Decisions (ADR-lite), resolved errors, review audits, and the consolidated index. What the team learned persists across sessions, searchable by tag, date, or theme.
Health
A fixed check set: plans, HANDOFF freshness, agents, commands, memory, git, config. Each failing check ships its own fix hint. Green means go. Red tells you exactly what to touch.
Git
Staging vs production diff, recent merges, branch state, and hook status. See what's ahead of main, what's queued for prod, and whether your guards are active.
Skills
Manage all your Agents Skills from one visual place with actionable buttons. Drop in a SKILL.md and it shows up here.
Commands
The full slash-command inventory: what each does, which are yours to edit, and a copy-ready prompt to create new ones.
Activity
See what's happening as you go: easy to browse active terminals, processes, git, agents and all the runtime at the tip of a finger.
Settings and updates from one place
The Config tab tunes cooldowns between ticks, audit preferences, agent persona, and external review behavior. Everything persists in a single config.json. Run agent-kit update to pull the latest commands, templates, and dashboard without losing your customizations.
Built with itself
Mission Kit is dogfooded daily. Every feature ships through the same /start-project to /git-prod pipeline it provides. Plans, reviews, triage, staging: the framework that builds your projects is built with the framework. Dogfood reports live under dogfood/ as real CI evidence.