International AI Builders Congress 2026 · CloudCamp BD: Lore Team tier will be showcased live on Event Day.See Tier Roadmap →

Your lore, structured.
Your AI, grounded.

Governed knowledge & Model Context Protocol (MCP) infrastructure for coding agents and enterprise teams. Return task-fit skills, deterministic verdicts, and traceable receipts — never blind RAG dumps.

# 1. Verify toolchain & install Lore Solo (Python 3.12 binary wheels)
uvx --managed-python --python 3.12 --no-build lore-solo==0.2.8 doctor
uvx --managed-python --python 3.12 --no-build lore-solo==0.2.8 install

# 2. Launch the local Bento setup dashboard & auto-connect your IDEs
lore-solo setup
lore-solo connect

Works across 9+ AI coding harnesses with distinct server keys

CursorVS Code CopilotGoogle AntigravityClaude DesktopClaude CodeOpenAI CodexDevin DesktopJetBrainsZed

Built for <developers> who want proof, not promises

Lore never settles human contradictions on its own. Every MCP call returns a typed verdict, cited rule IDs, and an auditable decision record.

Persona-Routed Skill Summaries

Instead of flooding the context window with an entire skill binder, lore-solo returns task-fit summaries and files to open, then logs a receipt via check_applied and log_decision.

lore-solo.lore/ OKFGraphify Map
{
  "schema": "lore.result/1",
  "server": "lore-solo",
  "decision_id": "dec-0042",
  "persona": "frontend-architect",
  "skills_served": ["bento-glass-tokens", "a11y-focus-ring"],
  "receipt_required": true
}

Add anything. Lore governs every decision

From local Markdown skills in Solo to multi-region PDF & OCR ingestion in Enterprise.

Step 01 detail
Dual-Stream Preservation: Digital PDFs extract at <1.5 ms/page with 0.00% CER. Legacy Bijoy/SutonnyMJ Bengali text is phonetically reconstructed into exact Unicode before indexing into Qdrant + FTS5.
Status: Drafts stay quarantined until human approval.

Built for one developer, scaling to regulated enterprises

Solo and Enterprise are live on dedicated branches. Team will be showcased on Event Day at the International AI Builders Congress by CloudCamp BD.

01

Project-as-Database (.lore/)

Plain Markdown & YAML skills under .lore/ plus an OS-wide shared persona library. Zero external database required.

02

Graphify Codebase Map

Optional tree-sitter binary-wheel codebase map (graphify-out/) with interactive viewer in the local setup dashboard.

03

5 Coding MCP Tools

get_context, get_skill, map_codebase, check_applied, and log_decision — focused on better-guided work.

Deterministic pipelines, verified end-to-end

Measured across LORE Enterprise's acceptance matrix, OCR benchmarks, and dual-stream ingestion suite.

Enterprise Acceptance Matrix100% PASS
Legacy Bengali Unicode Reconstruction (Bijoy / SutonnyMJ)100% Exact
Digital PDF Preservation Accuracy (pdfplumber, <1.5 ms/page)100% (0.00% CER)
Bounded GraphRAG Traversal Budget (2–3 hops, cycle-safe)≤ 100 ms timeout

Built for enterprise. Designed for control.

Airgap-capable, customer-deployed knowledge grid with fail-closed contracts and tamper-evident recovery.

L7–L0

9-Layer Governance

Pre-model permission pruning, temporal as_of queries, scoped waivers, and a 7-rank authority ladder that refuses to guess on ties.

RRF

Hybrid Qdrant + GraphRAG

Dense 384-dim embeddings + BM25/FTS5 merged via Reciprocal Rank Fusion, expanded over typed dependency & conflict edges.

SHA

Audit Lineage & DR

Immutable DecisionRecord & AuditEvent logs with trace_id links, Review Queue gating, and SHA-256 tamper-evident backup drills.

Give your AI structure and proof