Roadmap
What we're building now, and everything we've shipped to get here — newest first.
Public beta
The core application is functional and actively being developed. We are not yet recommending CodePlans for production workloads. Early adopters are welcome to self-host, test the hosted beta, and provide feedback.
Open Source Release
v0.6.1 · MIT · shipping weeklyCodePlans is publicly available on GitHub and shipping fast — now at v0.6.1, with the full 71-tool MCP server, five tracker integrations, releases, the asset record, the Asset Atlas, Product Wiki, and the collaboration loop (reviews, My Work, notifications) all in the open-source build. The repository includes a contributing guide, MIT license, and self-hosting docs. Clone the repo, follow the SQLite setup guide, and be running locally in minutes — 412 automated tests keep it honest.
View on GitHub →Timeline — newest first
Round-trip engineering & hosted tier
The honest remainder still being built
With the model, the Atlas, releases, the asset record, and now the collaboration loop all shipped, the outstanding work splits two ways: closing the loop between the model and real code, and hardening the hosted tier. Agents will propose reconciliation of the model against actual changes; humans stay in the approval seat.
- —Round-trip engineering — agents propose model ↔ code reconciliation
- —AI effort estimation and plan-delta insights (drafting already ships)
- —Scheduled sync + webhooks — today sync is on-demand
- —Hosted multi-org, SSO, and Stripe team billing
- —Analytics dashboard — plan velocity, throughput, per-product summaries
Collaboration loop
Responsibilities, reviews, My Work, and notifications
The most recent release. Responsibilities (code owners, architects, eng managers) route plan and spec reviews pinned to a revision, with anchored comments and @mentions. My Work turns those responsibilities into a personal inbox with role-based lenses. Notifications reach people in-app, by email, and in Slack — with admin rules and per-user opt-outs.
- —Plan & spec reviews pinned to a version; Open vs. Guided workflow
- —Anchored comments with @mentions
- —My Work inbox: Needs you, In flight, Watching + role lenses
- —In-app, email (Resend), and Slack notifications with opt-outs
Product Wiki & deletion safety
A readable knowledge base, and nothing disappears silently
Product Wiki turned asset pages and specs into a full-window, searchable document reader with contextual links and attribution. Alongside it, hard deletes across assets, products, tasks, work items, plans and releases were replaced with reversible archive/soft-delete tombstones, backed by a centralized audit log and role-based delete authorization.
- —Full-window asset pages and document readers, product-wide search
- —Reversible archive/soft-delete for assets and products — nothing referencing them breaks
- —Centralized audit log across UI and MCP mutations
- —Role-based delete authorization; creator/updater attribution surfaced everywhere
Layers & Native Specs
Architectural boundaries and versioned specs
Assets gained a layer field with Atlas layer columns and boundary-rule guidance for agents. Specs became first-class and versioned — associated with assets, plans, and work items, with pinned delivery receipts and legacy URL import for anything captured before.
- —Asset layer field, Atlas layer columns, move_asset refactoring
- —Versioned specs with asset/plan/work-item associations
- —Pinned delivery receipts on shipped work
- —Legacy spec URL import; consistent GFM rendering
Asset Atlas
A living map of the whole system
A live system map projected from your inventory — products as columns, dependency edges between assets, and lenses that recolor the map by health, debt, or delivery activity. Because it is derived, not drawn, it cannot rot. Grid and sortable table views sit alongside it.
- —Map, grid, and sortable table views of every asset
- —Health / debt / activity lenses
- —Dependency edges: depends on, integrates with, aggregates
- —Blast-radius hover — see what a change touches before you plan it
Asset Record
A capabilities register built from delivered work
A per-asset record of what each component actually does — assembled only from delivered work items, never hand-waved. Every capability carries its lineage back to the work item, plan, and release that shipped it. Removals become tombstones, so the record never goes quietly stale.
- —Capabilities graduated from resolved work items
- —Delivery lineage on every claim
- —Derived known-issues and debt sections
- —Tombstoned removals instead of silent deletions
Releases, Versions & Asset History
Delivery grouping and time-aware assets
Releases group the plans that ship together — a version of one asset or a coordinated multi-asset revision — with per-asset version stamps and release notes derived from delivered work. Every asset gained a version-structured history timeline and a curated design log authored by people and agents alike.
- —Releases group plans; per-asset version stamps
- —Release notes derived from delivered work items
- —Version ladder + derived asset history timeline
- —Design log — user- and agent-authored, attributed
- —Feature-flagged AI drafting (BYO Anthropic key), always human-approved
MCP Server & Tracker Integrations
CodePlans inside your AI-native editor
A full MCP server — now 71 tools over Streamable HTTP with read/write-scoped API keys — reachable from Claude Code, Cursor, Codex, GitHub Copilot, and Antigravity. Landing alongside it: five tracker integrations (GitHub, GitLab, Jira, Asana, Linear) with pull-only mirroring and strict provenance. All of it works in the OSS build.
- —71 MCP tools over Streamable HTTP; read/write-scoped API keys
- —Claude Code, Cursor, Codex, Copilot, and Antigravity
- —Five trackers, pull-only mirroring with strict provenance
- —PR auto-linking updates per-asset delivery status
- —Capture /docs into structured plans; setup snippets built in
Open Source Foundation
SQLite + pluggable backends
v0.2.0 decoupled CodePlans from Supabase, introducing an adapter pattern that lets the auth and database layers be swapped entirely via environment variables. This is the foundation for open-source, self-hosted distribution.
- —Auth adapter: Supabase Auth ↔ Auth.js credentials (JWT)
- —Database adapter: PostgreSQL ↔ SQLite (via libsql/Drizzle)
- —Separate SQLite and PostgreSQL migration tracks
- —HOST_MODE / REGISTRATION_ENABLED / BILLING_ENABLED flags
Foundation
Core application running on live data
The first production-ready cut wired a complete Next.js application to a live PostgreSQL database (Supabase) using Drizzle ORM. All core modules — authentication, products, assets, code plans, tasks, and team management — were connected to real data.
- —Full Drizzle schema: users, organisations, products, assets, plans, tasks
- —Supabase Auth with route-protection middleware
- —Server Actions for all create/edit/delete operations
- —Seed data and development workflow
Stay in the loop
Sign up for the beta to get updates as phases ship.