Roadmap

OIR development is organized in sprints. The full plan lives in ROADMAP.md at the repository root.

Current Sprint: MCP Security and Write Tools

We are hardening the MCP server and closing remaining content gaps:

• MCP authentication and rate limiting before any public HTTP endpoint
• Write tools (suggest_edit, submit_intake, resolve_debt_item) that open labeled PRs for human review
• GitHub App integration for AI-proposed changes
• Viacom v. YouTube case page and high-priority research debt resolution
• Technical expert pages for litigation consultation — Cerf, Vixie, Diffie, Zimmermann, Zittrain added; Kahn, Crocker, Berners-Lee remain in intake queue

MCP access model

Use case	How to connect	Public URL needed?
Cursor (local dev)	.cursor/mcp.json — stdio mode	No
Production server	oir-mcp on 127.0.0.1:8080	No (localhost only)
Remote AI agents (future)	HTTPS at /mcp/ with API keys	Yes — after auth ships

The MCP server is intentionally not public today. Current tools are read-only; exposing HTTP without authentication would allow unauthenticated querying and abuse of URL verification. See MCP_SERVER.md for details.

Recently Completed

Sprint 11 — Editorial workflow, Decap CMS, and MCP read path

• Moderation queue, taxonomy audit, GitHub labels, and scheduled report workflows
• Decap CMS at /admin/ with GitHub OAuth on production
• OAuth env-file pattern for server secrets (no credentials in shell history)
• MCP server with 9 read-only tools; Cursor project config (.cursor/mcp.json)
• Production oir-mcp service on localhost

Sprint 10 — Content depth

• Law firms (Perkins Coie, Wilson Sonsini, Cooley, Fenwick)
• Policymakers (Wyden, Paul, Khanna, Lofgren)
• Additional cases and topics (Google v. Oracle, Section 230, intermediary liability)

Sprint 9 — Public deployment

• Site live at openinternetresearch.com
• GitHub issue/PR templates, CODEOWNERS, editorial workflow docs
• Auto-deploy via webhook and RestartAndUpdateOIR.sh

Foundation in Place

• 50+ knowledge pages, bibliography, contacts, and generated indexes
• Metadata validation, link checking, relationship graph, handbook, and CI
• Tools for URL health, research debt, moderation reports, and label sync
• Documented publishing, OAuth, and MCP setup guides

Backlog

• Curated timeline narratives and advanced citation analytics
• Advanced outreach synchronization and contact lifecycle tracking
• Automated PDF rendering beyond browser print
• Advanced AI retrieval chunking and embeddings
• Zensical migration (when MkDocs successor is production-ready)