# mission-control

> The world's first Autonomous Product Engine (APE): AI agents research your market, generate features, and ship code as PRs. Convoy mode, crash recovery, cost tracking, 80+ API endpoints. Self-hosted v

- **URL**: https://www.freshcrate.ai/projects/mission-control
- **Author**: crshdn
- **Category**: Infrastructure
- **Latest version**: `v2.5.1` (2026-04-29)
- **License**: MIT
- **Source**: https://github.com/crshdn/mission-control
- **Homepage**: https://autensa.com
- **Language**: TypeScript
- **GitHub**: 1,931 stars, 417 forks
- **Registry**: github
- **Tags**: `aiagent`, `automation`, `openclaw`, `typescript`

## Description

The world's first Autonomous Product Engine (APE): AI agents research your market, generate features, and ship code as PRs. Convoy mode, crash recovery, cost tracking, 80+ API endpoints. Self-hosted via OpenClaw Gateway.

## Recent releases

| Version | Date | Urgency | Changes |
| --- | --- | --- | --- |
| `v2.5.1` | 2026-04-29 | High | ## Added - Repo Setup tab for Autopilot products with git, branch, GitHub API, PR metadata, Actions, workflow token, secret, and variable readiness checks. - User-fixable GitHub setup actions for workflow token permissions, Actions secrets, and Actions variables. - PR checks recovery in Build Queue with failed check classification, GitHub Actions reruns, and external check rerequests.  ## Fixed - Private repo validation now uses authenticated git access and branch confirmation. - Workspace isola |
| `v2.5.0` | 2026-04-14 | High | ## What's Changed  ### Fixed - **Per-task OpenClaw sessions** ([#99](https://github.com/crshdn/mission-control/issues/99)) — Dispatch now creates a unique session per agent + task combination instead of reusing a single session per agent. Previously, all tasks on the same agent shared one conversation session, causing context to accumulate until the model's context window was exhausted and the agent stalled. The `openclaw_sessions.task_id` column (already in schema) is now used for session looku |
| `v2.4.1` | 2026-04-12 | High | ## Community Bug Fixes  - **Autopilot model routing** — Provider models now route through `openclaw/default` with the original model in `x-openclaw-model`, fixing 404 errors on OpenClaw deployments. ([@Ahmedkasmi-dev](https://github.com/Ahmedkasmi-dev), [#109](https://github.com/crshdn/mission-control/pull/109)) - **AUTOPILOT_MODEL env var** — Removed hardcoded model override in description generation so the shared `AUTOPILOT_MODEL` config is respected. ([@aaronmeza](https://github.com/aaronmeza |
| `v2.4.0` | 2026-03-23 | Medium | Hat tip to [Nous Research](https://github.com/nousresearch) for these.  Closed-loop skill system where agents autonomously create, consume, and improve structured playbooks. Based on the Karpathy AutoResearch pattern — agents learn reusable procedures that compound over time.  ## How It Works  1. **Task completes** → LLM extracts 0-3 reusable skills from activities and deliverables 2. **Next dispatch** → Matched skills injected as primary instructions (keywords + role filtering) 3. **Age |
| `v2.3.1` | 2026-03-22 | Medium | Fixes startup errors from v2.3.0 merge artifacts.  ## Fixed  - **Schema syntax error on fresh DB init** — Missing `);` between `user_task_reads` and `product_program_variants` table definitions caused `near "CREATE": syntax error` on startup. Artifact from PR #89 merge resolution. - **Migration ID collision** — Migration 022 was previously applied as `error_reports` (since removed), causing the new `health_scores` migration and all subsequent migrations (023-027) to be skipped. Fixed migration r |
| `v2.3.0` | 2026-03-22 | Medium | Six new features, security fixes, and quality-of-life improvements.  ## Added  - **Idea Similarity Detection & Deduplication** — New ideas are compared against existing ones using text similarity. Ideas >90% similar to previously rejected ideas are auto-suppressed with a full audit trail. Remaining similar ideas get a warning badge on the card. ([PR #88](https://github.com/crshdn/mission-control/pull/88)) - **Operator Chat Overhaul** — Floating chat widget accessible from any page. Threaded conv |
| `v2.2.1` | 2026-03-22 | Low | Monitoring integration and database backup management.  ## Added  - **Health Check Endpoints** — `GET /api/health` returns unauthenticated summary or authenticated full-detail JSON covering DB integrity, gateway connectivity, agent status, task queue depth, research cycle freshness, and cost cap utilization. `GET /api/health/metrics` exposes Prometheus text format for Grafana/monitoring dashboards. ([PR #87](https://github.com/crshdn/mission-control/pull/87)) - **Database Backup API** — On-deman |
| `v2.2.0` | 2026-03-22 | Low | Swipe-driven preference learning (Karpathy AutoResearch pattern) and token usage tracking fix.  ## Added  - **Preference Learning** — Every swipe rebuilds a per-product preference model. Analyzes category approval rates, complexity preferences, impact score thresholds, tag patterns, and examples of approved/rejected ideas. The resulting `learned_preferences_md` is injected into research and ideation prompts, steering future cycles toward what the user wants. - **Preference Backfill API** — `POST |
| `v2.1.1` | 2026-03-22 | Low | Fixes a CHECK constraint failure that prevented ideas from being stored when the LLM returned a category not in the schema's allowed list.  ## Fixed  - **Ideation CHECK constraint failure** — LLM occasionally returns idea categories not in the schema's allowed set (e.g. `"analytics"`, `"design"`, `"marketing"`). Categories are now validated before insert, falling back to `"feature"` for unrecognized values. This was causing the entire ideation cycle to fail silently after ideas were parsed but b |
| `v2.1.0` | 2026-03-22 | Low | Server-side autopilot pipeline, toast notifications with one-click error reporting, and pending ideas badges.  ## Added  - **Server-Side Pipeline Orchestration** — The research → ideation pipeline now runs entirely server-side. "Run Now" fires a single request and returns immediately. Navigate away, close the tab — the pipeline keeps running. Multiple products run concurrently. - **Toast Notification System** — Global toast notifications surface errors, warnings, and status updates in real-time. |

## Citation

- HTML: https://www.freshcrate.ai/projects/mission-control
- Markdown: https://www.freshcrate.ai/projects/mission-control.md
- Dependencies JSON: https://www.freshcrate.ai/api/projects/mission-control/deps

_Generated by freshcrate.ai. Indexes github releases for AI-agent ecosystem packages._