CueMap
CueMap is a high-performance temporal-associative memory engine for fast, accurate, and explainable context recall. It turns natural-language queries into inspectable cues, intersects evidence, and returns source-backed memories for agents and other applications.
At a Glance
- AI coding-assistant developers
- Developers building agents and RAG systems
- Teams needing durable project or organizational memory
- Support, operations, and incident-response teams
- +2 more
AI Tools by CueMap
(1)
CueMap
AI Agent Memory Store in Rust
Discussions
No discussions yet
Be the first to start a discussion about CueMap
Latest News
CueMap v0.7.2 release published; SDK release notes add semantic modes, query embeddings, intent classification helpers, directory preview, and qint8/q4 MiniLM-L3 documentation.
CueMap v0.7.2 changelog: bundled local semantic reranking, intent classification, caller-provided vectors, edge semantic profile, snapshot recovery, and expanded coverage gates.
CueMap v0.7.1 added bundled tokenizer assets, Linux compatibility improvements, rustls-based HTTP TLS, consolidated production Docker builds, and health checks.
CueMap v0.7.0 introduced CuePacks, CueBridge artifacts, ordered/evidence recall modes, deterministic facets and intent routing, and removed the embedded web UI.
Products & Services
The core Rust temporal-associative memory store and HTTP/CLI engine, providing deterministic cue extraction, lexical and hybrid recall, project isolation, ingestion, persistence, and source-backed/grounded recall.
Python client for the CueMap engine, including synchronous and asynchronous clients, memory writes, recall, grounded recall, ingestion, project operations, backups, and intent classification.
TypeScript client wrapping the CueMap HTTP surface, with local/embedded-engine support and APIs for memory, recall, ingestion, backups, lexicon, and job operations.
An MCP server that automatically manages a local Rust CueMap instance and exposes project initialization, ingestion, status, memory lifecycle, aliases, lexicon administration, and ranked recall to coding assistants.
Market Position
CueMap positions itself as a deterministic, inspectable memory layer rather than a black-box semantic-memory service: lexical candidate discovery is bounded and evidence remains tied to original sources, while local MiniLM reranking is optional/bounded and requires no external model calls or network service. Its differentiators are project isolation, temporal and reinforcement signals, grounded recall, codebase ingestion, and a local MCP path for coding agents.
Leadership
Founders
Kaan Demirel
Software engineer focused on high-performance AI infrastructure and open-source projects; his GitHub profile says he is building CueMap and bridging probabilistic and deterministic approaches. He is the author/maintainer shown on the CueMap Python package and the owner of the cuemap-dev GitHub organization.
Executive Team
Kaan Demirel
Founder / maintainer
Software engineer building high-performance AI infrastructure and open-source projects; listed as the CueMap package author and the sole member shown for the CueMap Hugging Face organization.
Founding Story
CueMap began as a Rust engine prototype for in-memory memory storage, basic tokenization, and exact-match scoring (v0.1.0, August 2025). Its stated initial vision was a high-performance temporal-associative memory store that could provide deterministic, inspectable evidence for dynamic contextual retrieval rather than opaque generated summaries.
Business Model
Revenue Model
CueMap is distributed as an open-source engine and client packages under source-available/MIT-licensed components: the engine repository uses Business Source License 1.1 with a database-service restriction, while the Python and TypeScript SDKs and MCP package state MIT licensing. The public materials emphasize local self-hosted use through CLI, SDK, MCP, HTTP, and Docker rather than a hosted subscription.
Target Markets
- AI coding-assistant developers
- Developers building agents and RAG systems
- Teams needing durable project or organizational memory
- Support, operations, and incident-response teams
- Applications requiring local, explainable, privacy-conscious retrieval
- Constrained or edge-device deployments
- Coding-agent memory and project-scoped codebase recall
- Long-running personal and workplace assistants
- Support and incident investigation
- Time-sensitive personalization
- Agent workflow, tool history, and state recovery
- Decision and organizational memory