Skip to main content

Blueprints

Blueprints are templates. Agents are instances. A blueprint avoids repeating config across similar agents.

Blueprint: Backend Developer
  Agent: Backend Implementer - Rust
  Agent: Backend Implementer - TypeScript
  Agent: Backend Implementer - C#

Blueprint fields​

  • default skills
  • default tools
  • default permissions
  • default memory policy
  • default runtime image

Why blueprints​

Without blueprints, three "backend implementer" agents repeat the same skill list, tool set, and permission matrix. A blueprint centralizes the defaults; each agent overrides only what differs (model, language-specific skills).

Relationship to agents​

When you create an agent from a blueprint, Nexus copies the blueprint's defaults into the new agent document, then lets you override fields. Changing a blueprint does not retroactively mutate existing agents — instances are independent documents once created (you may add an explicit "re-sync from blueprint" action later).

Stored shape​

{
  "_id": "blueprint_backend_developer",
  "name": "Backend Developer",
  "default_skills": ["git-branch-workflow", "api-test-writing"],
  "default_tools": ["git", "shell", "plane-comment"],
  "default_permissions": { "can_commit": true, "can_open_pr": true, "can_merge": false },
  "default_memory_policy": {
    "read_scopes": ["global", "project", "agent_private"],
    "write_scopes": ["project", "agent_private"],
    "auto_save": true
  },
  "default_runtime_image": "nexus-agent-runner:latest"
}