Open-source EvoOntology lets agents build and evolve a data ontology via MCP
TheTuringPost · x · 2026-09-26
Researchers at Renmin University open-sourced EvoOntology, a self-evolving ontology layer that helps AI agents understand business data by connecting concepts to actual data sources and updating the guide based on task history.
How it works: a builder agent examines real tasks and data sources to create the initial ontology — definitions, field mappings, relationships, and business rules with evidence from the data. It has three layers: Schema (representable concepts), Content (definitions, data links, rules), and Tools (agent search). Agents query the ontology via MCP tools rather than prompts. After tasks, the system proposes ontology changes, keeping only those that improve the score.
Gains: +17.8 points accuracy on multi-source research, +1.9 on insight mining, +7.4 query accuracy and +8.6 execution-efficiency on data retrieval. Ships with Claude Code and Codex plugins for building, updating, and visualizing the guide.
Related event: RUC Open-Sources EvoOntology for Data Agents(3 posts)→
More from coding & agent
- Repurpose your old M1/M2 Mac as a local test server with this open-source runner script — jasonkneen · 2026-09-26
- DHH says 37signals went "pencils down": 150k lines in August with agents vs 30k/year by hand — Recent-Tangerine2745 · 2026-09-26
- Harness engineering: the key discipline for building reliable AI agents — techNmak · 2026-09-26
- Feeding PTX ISA to LLMs: Converting Figures to ASCII Art with Gemini — bingxu_ · 2026-09-26
- Merge Claims Its MCP Connectors Match or Beat Public MCPs, Cut Tokens 24% — shensi · 2026-09-26
- Define Dimensions First: Hamel Husain's Method for Synthetic Eval Data — HamelHusain · 2026-09-26