Sector brief · April 2026
Primary profile: Knowledge base (/k/)
Enterprise knowledge management
Enterprises finally agree on RAG. They disagree on how much it should cost. Serialization choices show up on the invoice the same way any other infrastructure decision does.[4][1]
Market context
Landbase-style surveys place knowledge-management venture activity in the multi-billion range. Buyers want assistants that cite internal docs without exfiltrating everything to a model vendor.[3]
Practitioner threads on TOON vs JSON repeat the same complaint: repeated keys and whitespace in JSON tax every high-frequency tool loop.[4][2]
Structural gaps
Where incumbents sit
Glean indexes everything behind enterprise SSO. LuMay orchestrates departmental agents. LangChain is the assembly line.[5][6][7] The wedge underneath all three is cheaper, schema-stable bytes at retrieval time.
Building on more.md (implementation map)
Knowledge base profiles (`/k`) as the governed corpus
Hang curated corpora, playbooks, and retrieval policies off `/k` entities so assistants resolve a stable handle before they pull chunks.
Ingest to Markdown
Run PDFs and spreadsheets through docling-style workers so structure survives conversion.[3]
TOON at the data plane
Honor `Accept: text/toon` for flat tables so headers are declared once per response.[2][4]
MCP for developer surfaces
Expose the platform MCP server so Cursor-class clients pull governed chunks without a parallel integration stack.[7]
Competitive context
Names below appear in third-party sources listed under References. The contrast column sketches where an agent-first build on more.md sits relative to each player. Complement in some layers, substitute in others.
| Player | Position in market | Contrast with more.md-shaped build |
|---|---|---|
| Glean | Enterprise search with broad connector coverage. | Central index vs payload compression at the more.md edge. |
| LuMay AI | Cross-department agent orchestration. | Application layer vs serialization wins under any orchestrator. |
| LangChain | RAG and workflow framework. | Routing logic vs shrinking per tool-call payloads. |