The planning and memory layer for AI-assisted software development
Structured checklists that give your AI agents the context to plan software correctly, and a searchable knowledge base that remembers every decision so nothing gets re-litigated or re-explained.
Your AI tools are only as good as the context you give them
Teams adopting AI-assisted development face a structural problem: AI agents are powerful but context-blind. Without clearly defined requirements, architecture decisions, security constraints, and acceptance criteria, even the best AI models generate code that misses edge cases, skips critical concerns, and drifts from what was actually needed — and every new session starts from zero, with no memory of what the team already decided.
Engineering Framework solves both halves of that problem. A curated checklist of planning artifacts — with definitions and fill-in templates — gives your team and your AI agents a shared, structured way to plan. A queryable knowledge base captures every decision as it's made and resurfaces it automatically, so prior art, past rationale, and existing answers are always one prompt away instead of being re-explained or accidentally contradicted.
Four capabilities that set your AI up to succeed
Built from the ground up for engineering teams that use AI coding agents
Structured checklists, definitions & templates
Browse a curated library of engineering planning artifacts — security requirements, SLOs, ADRs, rollback plans, threat models, user stories, and more. Every item ships with a definition that explains its purpose and a fill-in template ready for AI generation, linked back to your Git repository as it evolves.
A knowledge base that remembers your decisions
Supported artifact types are automatically captured as searchable drafts and refined through review. Ask a question in plain language and get back the prior art, past rationale, or existing answer — scored for confidence — instead of starting from scratch.
AI agents connect natively via MCP
Two always-on MCP servers put planning tools and knowledge search directly in your AI agent’s hands — generate and review artifacts, check planning health, and query prior decisions without leaving your terminal.
Fine-grained permissions
Role-based access control at the organization, department, and project level — enforced consistently on the UI and on every MCP tool call. Admins, tech leads, members, and guests each get exactly the access they need.
Everything you need to plan and remember with AI
A complete platform for structured, AI-native engineering planning and institutional memory
Checklist library with definitions & fill-in templates
Every planning artifact — SLOs, rollback plans, threat models, success criteria, vendor requirements, ADRs, and more — ships with a definition and a markdown template optimized for AI generation.
Ask the knowledge base in plain language
Search across every past decision, artifact, and reviewed document by meaning, not keyword. Get ranked, confidence-scored answers your AI agent can act on directly.
Completed work becomes searchable automatically
Supported planning artifacts flow straight into the knowledge base as drafts, ready for review — no separate step to remember to write things down.
A review workflow you can trust
Published entries are approved by a domain reviewer first. Health flags surface anything that has drifted, gone stale, or never been reviewed, so you always know what to trust.
Two MCP servers, one AI agent
A planning server exposes start_planning_session, generate_artifact, and review_artifact; a knowledge base server exposes search and retrieval. Both connect natively to Claude Code and other MCP-compatible tools.
Auto-generated config for every project
Each project generates a ready-to-use setup snippet that scopes your AI agent to the correct organization, project, and MCP server endpoint — no manual wiring.
Git repository integration
Link projects to Git repositories and publish generated planning artifacts alongside your code. Engineering context lives where developers work.
Planning health & phase tracking
Track artifact completion across configurable project phases. Know exactly what is done, what is in progress, and what is still missing before development begins.
Fine-grained authorization — UI and MCP enforced
Every permission check runs through relationship-based access control. Roles, organizations, departments, and project memberships are enforced uniformly across the web UI and all MCP tool calls.
Multi-tenant organization & department management
Support multiple organizations, nested departments, and distinct project teams — each with their own members, roles, and permission boundaries.
Configurable project phases
Define the planning phases that match your engineering process. From discovery through architecture, implementation, and launch readiness — track progress at each stage.
Built for AI-assisted engineering teams
Whether you're an individual contributor or running a multi-team organization
Engineers using Claude Code
Give your AI assistant the structured context it needs — requirements, constraints, acceptance criteria — before generating a single line of code, plus instant recall of every past decision. Invoke planning and knowledge tools directly from your terminal session.
Tech Leads
Ensure your team works from a consistent set of architecture decisions, security requirements, and success criteria before AI-assisted sprints begin. Catch planning gaps early, not in code review.
Engineering Managers
Gain visibility into planning completeness across all projects, departments, and teams. Set organization-wide standards once and enforce them through role-based access control.
AI-first engineering teams
Purpose-built for teams that rely on AI agents for day-to-day development. Structured planning artifacts become the shared context that keeps humans and AI aligned on what needs to be built.
Start planning with AI at your side
Give your team and your AI agents the structured context they need. Curated planning artifacts, native Claude Code integration, and fine-grained permissions — all in one place.