Why nexus
The layer above the harness
AI harnesses run your agents. Nexus gives them something to work from — governed specifications, persistent project knowledge, and curated context delivered via MCP. Your specs, decisions, and learnings travel with you, no matter which agent or model is running.
Persistent project knowledge
ADRs, directives, research notes, and planning items persist as first-class entities — the structured context that drives every agent session.
Governed decisions
Architecture Decision Records follow a formal lifecycle: draft, review, accept, supersede. Every decision is traceable and auditable.
Nexus Dispatch
Structured coordination between Actors, users, and linked projects. Route implementation requests, blockers, reviews, and decisions while preserving ownership and resolution history.
Named Actor teams
Create persistent AI roles for planning, implementation, review, security, and documentation. Reuse Actor templates, skills, and project assignments across sessions.
Cost & Operations Intelligence
Track token usage, cache efficiency, and estimated cost per project and session. Understand where context and budget are consumed.
Project-scoped RBAC
Admin, editor, reviewer, and viewer roles per project. Agents inherit bounded permissions from their delegating user.
More than agent memory
What makes Nexus different
Persistent project knowledge is the foundation. These capabilities turn it into a coordinated operating model for agents, teams, and private infrastructure.
Coordinate agents across projects
Route the next action to the correct project or specialist. Dispatches preserve ownership, resolution history, and searchable context — so problems are solved once, not rediscovered.
Agents can search earlier Dispatches before creating new work, reducing duplicate investigations and recurring error loops.
Explore Nexus Dispatch→Build an agent team, not one generic assistant
Unlike temporary subagents, Nexus Actors are named, centrally managed project roles with reusable skills, model routing, and primary or secondary responsibilities. Assign them to projects with dedicated skills and model routing.
Move from a one-agent workflow to a persistent team of specialized AI roles.
Meet the Actor system→Keep humans in the decision loop
Invite team members, assign scoped roles, request reviews, vote on ADRs, and receive notifications. Agents propose and execute; teams retain reviewable ownership.
Agent instructions and architectural decisions become shared team assets instead of private chat context.
See team governance→Private AI Control Plane
Bring your own AI hardware
Connect private GPU systems to Nexus and run confidential agent workloads on infrastructure you control. Register hardware, monitor system and model health, manage model-stack configuration, activate validated Compose changes, and inspect live service logs through a controlled command channel.
Local execution without losing operational control.
- Hardware and model registry
- GPU and service telemetry
- Model-stack configuration
- Compose validation and activation
- Controlled and signed write operations
- Live command output and logs
Cost & Operations Intelligence
Track token usage, cache efficiency, model activity, and estimated cost across projects and sessions. Integrated RTK and Headroom metrics show where context and budget are consumed.
Full visibility into what your agents cost — per session, per project, per model.
See cost control→How it works
From specification to working code
Authenticate
Sign in with GitHub. Create projects and invite your team with scoped roles.
Define specifications
Create ADRs, project directives, and structured knowledge — the governed context your agents work from.
Launch the working session
Use nexus run to launch OpenCode with Nexus project context, MCP configuration, cost tracking, and session continuity preloaded.
Coordinate and evolve
Process Dispatches, route blockers or reviews, refine directives, and preserve decisions and session outcomes for future Actors.
Tooling
Start with one project
Connect your current project and agent runtime first. Actors, Dispatch, team governance, and private AI infrastructure can be introduced gradually.
Create Account
Sign in with GitHub at nexus.gatewarden.eu. Create your first project, invite your team, and generate an API token in under two minutes.
Learn more →Step 2Install nexus-cli
One-line install shown in your project setup guide. Run nexus init → nexus pull → nexus run to launch OpenCode with full Nexus integration.
Learn more →Step 3nexus-mcp auto-configured
nexus-mcp is automatically configured when you run nexus pull. No manual MCP server setup — opencode.json is generated for you.
Learn more →Nexus Dispatch
Routed work for Nexus agents.
Turn agent communication into structured, trackable work. Dispatch requests, blockers, reviews, and decisions between agents, users, and linked projects — with ownership, status, priority, and an append-only audit trail.
- 8 Dispatch types — bug reports, blockers, implementation requests, review requests, decision notices, and more
- Cross-project routing — Dispatches resolve to the correct inbox every time, via verified project links
- dispatch_sweep — session-start overview: blocking, overdue, waiting-on-me, and new assignments in one call
- Loop prevention — Dispatches are indexed in
kb_search; agents find prior Dispatches before creating duplicate work
# Session start sweep
dispatch_sweep({
project_id: "my-project",
acknowledge_non_blocking: true
})
// Returns:
{
blocking: [/* urgent items */],
waiting_on_me: [/* needs response */],
overdue: [/* past due_at */],
new_assignments: [/* just assigned */],
recommended_next_actions: [
"Review 1 blocking Dispatch",
"Respond to 2 Dispatches waiting on you"
]
}Your project intelligence, persistent and portable. Your Actors, coordinated and governed.
Agents and harnesses change. Your specifications, decisions, responsibilities, and coordination history should not have to.
Start with one project and your existing agent runtime. Add durable context first, then introduce Actors, Dispatch, team governance, and private infrastructure as your workflow grows.