Full-body literal search
- Keyword and literal search now match indexed document body text via hierarchical chunks — not just titles and previews
- Literal document hits include a highlight snippet and a chunkId that works with Deep Search and the /v1/pd endpoints
- Dashboard keyword results render highlight snippets with the matched query bolded
- Documents over 4 MB now index their first 4 MB for search instead of failing
Reasoning over your documents
- New `reason` MCP tool and `POST /v1/reason` endpoint
- Synthesized, cited answers composed across your documents — with citations, gaps, and conflicts
- 29 MCP tools and 30 REST operations total
Agent Readiness v2
- Enhanced agent discovery surfaces
- OpenAPI 3.1 specification published
- Agent Skills index and A2A agent card
MCP Server v2.0
- OAuth 2.1 PKCE flow for ChatGPT and MCP clients
- 29 MCP tools with full CRUD, search, reasoning, and progressive disclosure
- Rate limiting with Upstash Redis sliding window
- API key authentication with bcrypt verification
Chrome Extension 1.0
- One-click webpage capture from any site
- Automatic content extraction and processing
- Direct save to Context Repo library
Deep Search & Progressive Disclosure
- Hierarchical chunk navigation for large documents
- Session-based search deduplication
- Vector similarity with 1536-dimension embeddings
Collections
- Collection-based organization with hierarchical structure
Documents & Version History
- Document management with 75+ file format support
- Full version history with visual diffs
- One-click version restoration
- Web scraping via Firecrawl integration
Initial Launch
- Prompt template library with variable interpolation
- Semantic search across all content
- REST API with 30 endpoints
- Claude Desktop MCP integration