Specification-driven execution
Link every task and note to the specification and keep scope traceable from intent to done.
Built for teams that ship continuously
Run delivery from one source of truth, not scattered tools.
ConstructOS connects project intent, specifications, tasks, notes, rules, and AI execution in a single operating layer. Your team keeps context with vector plus graph retrieval, moves faster, and keeps an audit trail. Team Mode gives multi-agent delivery a clear operating model, Git Delivery keeps execution tied to repository evidence, and Docker Compose keeps self-hosted rollout straightforward.
- Self-Hosted Only
- Codex + Claude Code + OpenCode
- Audit Trail Built In
- Vector + Graph Retrieval
- Jira + GitHub MCP
- Live Event Storming
Core capabilities
Everything needed to execute with context
Project memory
Keep project rules, notes, references, and decisions attached to the work where teams need them.
MCP-powered automation
Use MCP tools and automation workflows with first-class Jira MCP and GitHub MCP integrations, plus support for additional MCP connectors.
Hybrid retrieval (vector + graph)
Use semantic vector search plus graph traversal to surface dependency-aware context with precision.
Event Storming + DDD mapping
Keep bounded contexts, aggregates, commands, domain events, policies, and read models connected to day-to-day artifacts in one living diagram.
Live workspace visibility
See changes as they happen with real-time notifications and activity updates.
Rules and governance
Make operational rules explicit so automation runs through clear boundaries instead of hidden convention.
REST + SSE delivery surface
Drive the platform through structured APIs and live server-sent events for updates, task changes, and notifications.
Live agent chat
Work with Codex, Claude Code, or OpenCode in real time, with project context, specs, tasks, and notes available in one chat surface.
Platform modules
Team Mode leads the operating model
ConstructOS is currently in beta, with the platform centered on team orchestration, verifiable delivery, and predictable self-hosted deployment.
Primary workflow
Team Mode
Coordinate multi-agent delivery through explicit role coverage, task routing, structured status flow, and kickoff automation. Team Mode is the fastest way to move from planning into controlled execution.
- Clear lead, developer, and QA workflow boundaries
- Kickoff and dependency-aware execution flow
- Status semantics designed for reliable automation
Delivery contract
Git Delivery
Tie execution back to repository state, verifiable changes, and QA evidence so delivery automation can be audited instead of guessed.
- Repository-aware execution boundaries
- Evidence-backed QA and merge flow
- Safer automation with explicit delivery checks
Deployment path
Docker Compose
Keep deployment simple with a self-hosted runtime model built around Docker Compose for controlled environments and repeatable rollout steps.
- Self-hosted only deployment model
- Operationally clear environment setup
- Predictable path from package to running stack
System shape
Architecture that engineers can reason about
Deterministic write path
Commands append events, projections update read models, and retrieval combines vectors plus graph context so every state transition stays traceable. Event Storming links are inferred asynchronously as a secondary projection with review status controls.
UI --REST/SSE--> API
API --append--> Event Store
Projections --> SQL + Graph
Event Storming --> Derived Graph Projection
Review API --> Link Status (candidate/approved/rejected)
Context API --> Vector Index + GraphAI execution with guardrails
MCP tools and automation runners operate through explicit command boundaries and policy checks.
Agent --> MCP Tools
MCP --> Domain Commands
Command IDs --> Idempotent MutationsHow it works
From planning to execution in three steps
01
Define intent
Create projects, specifications, and rules that become the source of execution context.
02
Run Team Mode
Route work through Team Mode so lead, developer, and QA responsibilities stay explicit while execution keeps moving.
03
Deliver and deploy safely
Use Git Delivery and Docker Compose to move from verified changes to self-hosted rollout with controls.
FAQ
Common questions
Is ConstructOS only for software teams?
No. Any delivery team with structured work, dependencies, and knowledge-heavy workflows can use it.
Can we self-host?
Yes. ConstructOS is self-hosted only and designed for teams that want deployment and operational control.
Do you support AI operations safely?
Yes. Automation flows use controlled command execution, idempotency patterns, and audit trails.
How does the Event Storming diagram stay up to date?
Artifact changes trigger asynchronous analysis that updates Event Storming plus DDD links in the knowledge graph. Teams can manually approve or reject inferred links.
Do you support vector search?
Yes. ConstructOS supports semantic vector search combined with graph context traversal.
Which MCP integrations are highlighted?
Jira MCP and GitHub MCP are highlighted integrations, and ConstructOS also supports additional MCP integrations for broader workflows.
What is Team Mode?
Team Mode is the structured orchestration layer for multi-agent delivery, with explicit roles, status flow, and execution gates.
How does Git Delivery fit in?
Git Delivery keeps automation anchored to repository state, verifiable changes, and delivery evidence instead of loose task updates.
Why Docker Compose?
Docker Compose is the deployment model for running ConstructOS in your own environment with a predictable self-hosted stack.
Are execution runtimes optional?
No. ConstructOS currently supports Codex, Claude Code, and OpenCode as execution runtimes for core automation workflows.
Is there a hosted cloud app?
No. ConstructOS is self-hosted only and deployed through Docker Compose in your own environment.
Is the product in beta?
Yes. ConstructOS is currently in public beta.
How quickly can a team onboard?
Most teams can set up their first production workflow in under one day with starter templates.
Scale delivery without losing context.
Use ConstructOS as your execution layer for reliable output, clearer ownership, and safer automation.