The execution layer that connects AI agents to context.
LatticeEngine delivers governed context to AI agents through a retrieval API, permission-aware context assembly, and agent work packet delivery. It routes every request through P1 → P2 → P3 tiers, writes back verified outcomes to memory, and provides an evaluation harness for testing.
Your AI agents retrieve context blindly.
Most AI systems retrieve context through a single vector similarity search. They cannot distinguish between a verified fact and a draft suggestion, between an authorized access and a permission violation, or between a current document and a stale one. Every retrieval looks the same — a semantic approximation with no provenance, no tier precedence, and no trace.
LatticeEngine routes every retrieval through the right memory tier with the right permissions. It assembles context with full provenance, delivers agent work packets with constraints and acceptance criteria, writes verified outcomes back to memory, and provides an evaluation harness so you can test context quality before it reaches production.
From context request to governed agent delivery.
AI agents request context through the retrieval API, MCP server, or context pack export. Every request carries a role, query, scope, and token budget.
LatticeEngine routes requests through P1 → P2 → P3 in strict precedence order. Deterministic tiers are always consulted before semantic fallback.
Permission-aware context assembly scopes and packages context for the requesting agent. Unauthorized atoms are excluded regardless of semantic relevance.
Agent work packets and assembled context are delivered with full provenance, confidence scores, tier sources, and trace IDs. Every atom is attributed to its origin.
Verified agent outcomes are written back to memory as candidates for promotion. Completed work becomes future context, closing the improvement loop.
Governed context delivery for every AI agent you run.
Four interfaces. Every AI platform you already use.
LatticeEngine does not replace your AI assistants, coding agents, or copilots. It supplies governed context to them through four integration surfaces — retrieval API, context pack export, MCP server, and memory writeback.
# Retrieval API
POST /v1/context/retrieve
body: { role, query, scope, max_tokens }
→ returns: { atoms[], pack_metadata, trace_id }
# Context pack export
GET /v1/packs/{pack_id}/export?format=mdx
→ returns: assembled context pack in MDX, JSON, or Markdown
# MCP server
mcp://contextlattice.local/v1
tools: context.retrieve, atom.lookup, packet.next
# Memory writeback
POST /v1/memory/candidate
body: { session_id, outcome, candidate_atom }
→ returns: { candidate_id, review_status }