{
"$type": "site.standard.document",
"bskyPostRef": {
"cid": "bafyreif45yctqynf25dqjo77pchlceyviywrlylxfztg53mscavb6fb2ye",
"uri": "at://did:plc:25rdn5elo5izoxrmtis34zuk/app.bsky.feed.post/3mogxql54fia2"
},
"coverImage": {
"$type": "blob",
"ref": {
"$link": "bafkreigjqsrbc2pmtevnyif5zz7642lmf52qx6p32zsjeq6gltrllpthhi"
},
"mimeType": "image/webp",
"size": 64864
},
"path": "/aubakirovarman/im-building-cortexdb-an-agent-native-context-database-for-ai-agents-59fe",
"publishedAt": "2026-06-16T23:02:22.000Z",
"site": "https://dev.to",
"tags": [
"rag",
"agents",
"rust",
"database",
"https://github.com/AubakirovArman/CortexDB"
],
"textContent": "# I'm building CortexDB — an agent-native context database for AI agents\n\nMost modern RAG systems work like this:\n\n 1. Split documents into chunks\n 2. Generate embeddings\n 3. Store them in a vector database\n 4. Retrieve top-k similar chunks on query\n 5. Send them to an LLM\n\n\n\nIt works for simple use cases. But as AI agents become more autonomous and complex, a clear problem appears:\n\n> Agents don’t just need similar text chunks.\n>\n> They need **bounded, permission-safe, evidence-aware, and verifiable context**.\n\nThis is why I started building **CortexDB**.\n\n**GitHub:** https://github.com/AubakirovArman/CortexDB\n\n## What is CortexDB?\n\n**CortexDB** is a single-node, agent-native context database. Its main goal is to compile **ContextPacks** — structured, citation-rich, token-budgeted bundles of context for AI agents.\n\nInstead of returning raw chunks, it returns a ready-to-use package that includes:\n\n * Source citations\n * Explanation of why each piece was selected\n * Token usage information\n * Anomaly and conflict detection\n * Permission and scope awareness\n\n\n\n## Key Features\n\n * **ContextPack** — structured output format with citations and token control\n * **VERIFY FACT** — deterministic fact verification (including numerical conflicts)\n * **AQL** — custom declarative query language designed for agents\n * **Tool Registry** + **Typed Knowledge Graph**\n * Durable single-node storage (WAL + MVCC)\n * Published SDKs for **Python** , **TypeScript** , and **Rust**\n\n\n\n## Example: ContextPack\n\n\n json\n {\n \"token_budget_tokens\": 4000,\n \"estimated_tokens\": 2500,\n \"truncated\": false,\n \"citations_required\": true,\n \"cells\": [...],\n \"anomalies\": [...]\n }\n",
"title": "I'm building CortexDB — an agent-native context database for AI agents"
}