For agents

Public vs protected

What is available anonymously and what requires authentication on the Blueprint MCP endpoint.

Public tools, no auth required

Read-only retrieval tools are fully anonymous: principles.list, clusters.list, principles.get, clusters.get, examples.get, principles.search, examples.search, and assets.list. signals.feedback is also open to anonymous callers, but only when the user explicitly asks to leave feedback.

Protected tools, authentication required

Personalised tools require a valid Bearer token: me.learning_path and me.coaching_context (any authenticated plan). architect.validate, signals.report, and team.summarize require Pro or Teams plan. me.add_evidence requires course enrollment. Handoff tools (handoffs.operator, handoffs.partnership, handoffs.agency) require Pro or Teams plan.

Access model

Pass an Authorization: Bearer <token> header to unlock protected tools. Public retrieval tools continue to work with or without auth, authentication never degrades the read-only surface. Protected tools return a structured error if the plan does not cover them, not a 401. This is expected plan-gating behavior, not a connection failure.

Public · Read-only

Public tools

These tools are public and read-only. No authentication required.

principles.list(cluster?)
clusters.list()
principles.get(slug)
clusters.get(slug)
examples.get(slug)
principles.search(query, limit?)
examples.search(query, principle_ids?, difficulty?, library?, limit?)
assets.list()

Authenticated · Pro

Authenticated / Pro tools

They appear in the live registry, but require authentication and are not part of the public read-only promise.

me.learning_path()
me.coaching_context()
architect.validate(implementation_context, focus_area?, task?, language?, repository?, files?, goals?, example_limit?)
me.add_evidence(course_slug, stage_id, note)

me.coaching_context() is designed for a single context injection per session, call it once at the start and pass the payload downstream. Polling it on every turn doubles your input tokens without adding new information.

Choose the path that matches your plan

Keep the information architecture honest: public doctrine access is available now, while Pro is the practitioner layer for protected reports, saved evidence, authenticated MCP, and recurring next actions.

Available now

Basic / Free: public doctrine access for local agent setups

Use downloadable rules, prompt packs, and the read-only public MCP endpoint to expose principles and examples inside Claude, Codex, Cursor, Windsurf, GitHub Copilot, Gemini CLI, DeepSeek, Qwen, and other agent-aware clients without account setup.

Download Claude skill packs, Cursor and Windsurf rules, Copilot instructions, Gemini context, Markdown, JSON, and llms.txt
Call the public read-only MCP tools
Right starting point for solo exploration and content discovery

See pricing context →

Next increment

Pro: protected personal validation and deeper retrieval

Pro is the practitioner workspace where MCP and skills turn real workflow evidence into protected readiness reports, saved findings, authenticated coaching context, and recurring next actions.

Protected validation against the principles
Sharper example retrieval with freshness metadata
Built for individual practitioners applying the doctrine to product and code

See pricing context →

Later phase

Teams: shared governance and agent visibility

Team-level agent workflows come after the protected layer is proven. That path is about shared audit visibility, private overlays, and organisational consistency, not a separate handbook.

Shared audit visibility for agent use and validation activity
Private overlays and team-level audit visibility (in development)
A2A agent-to-agent: production agents (LangGraph, CrewAI, AutoGen) call Blueprint coach as a peer via A2A protocol, governance runs automatically in every agent pipeline (coming soon)
Only after the hosted personal path proves real demand

See pricing context →

How should a team use this page in the larger site?

Use it after the doctrine and examples are clear enough to operationalize. Principles define the standards, examples provide inspectable evidence, and this page distributes both into agent tooling.

Also in this section

MCP setup Trust and data handling Start here