Kerneta
Kerneta builds memory-management and shared-processing systems for AI. Its first product, DaiDocs, is an open .dai file format that stores an AI assistant's long-term memory as readable plain-text files owned by the user, portable across models and assistants.
At a Glance
- Developers and AI-agent users
- Startups and enterprises building on MCP or AI assistants
- Teams that need shared, governed AI memory
- Researchers evaluating long-context and memory systems
- +1 more
AI Tools by Kerneta
(1)
daidocs
Local AI Memory File System
Discussions
No discussions yet
Be the first to start a discussion about Kerneta
Latest News
V4.4n32 launch release published on GitHub; added project-scoped memory, automatic session saving and Claude subscription mode.
Reading protocol update: grep the manifest rather than loading it whole.
DaiDocs V4.4n32 launch release published with the reference engine, MCP server and Python reader package.
Kerneta launched the open .dai AI-memory format and DaiDocs, publishing its benchmark evidence and Apache 2.0 repository.
Products & Services
An open Apache 2.0 plain-text memory format with Identity (YAML), Understanding (JSON) and Content (verbatim text) zones; one file per conversation, with searchable JSONL indexes and byte-exact originals.
The reference engine that converts conversations into .dai files and retrieves the smallest relevant context using manifest, understanding and segment-level retrieval.
An MCP server exposing save_memory, recall_memory, list_memories, read_memory, declare_project and brief_parent to compatible assistants and agents.
A Python package and daidocs command that reads .dai stores and drives the Node engine.
Market Position
Kerneta positions .dai as an open, local-first memory format rather than a hosted memory service: files remain on the user's disk, are readable with ordinary tools, and work across vendors. Its official benchmark page places .dai second on LongMemEval-S at 83.00%, behind Mastra Observational Memory at 84.80% but using about one-third of the context, and ahead of Supermemory, TiMem, Zep and Feather in the comparable table. The site also contrasts it with built-in Claude/ChatGPT memory and vector-database/server-based memory products.
Leadership
Founders
Amin Rigi
Founder of Kerneta and active director of Siro Robotics Ltd. His doctoral research at the University of Edinburgh focused on biosensing and wireless, battery-free RFID sensing systems. He built the world's first lifeguard robot in 2012, has won 15 national and international awards, and co-founded Eyesight Electronics, an eye-health company. At Kerneta he created the .dai format and built its retrieval engine.
Ali Munir
Co-founder. He has a Master's degree in AI and specializes in applying AI/ML to embedded systems, real-time systems and IoT architecture. His work includes training and deploying models on resource-constrained hardware, environmental sensing, edge intelligence on microcontrollers and IoT devices, and leading embedded-systems engineering across production hardware platforms.
Executive Team
Amin Rigi
Founder
University of Edinburgh doctoral researcher in biosensing and wireless battery-free RFID sensing; builder of a 2012 lifeguard robot, co-founder of Eyesight Electronics, and creator of the .dai format and retrieval engine.
Ali Munir
Co-founder
AI/ML and embedded-systems engineer with a Master's degree in AI; background in real-time systems, IoT architecture, edge AI, environmental sensing and production hardware compliance.
Founding Story
Kerneta began as an internal fix at Sirotics: long-running AI programs kept losing their own history and teams repeatedly rebuilt fragile memory layers. Rather than solve the problem separately for each client, the team built an open, vendor-agnostic file format and spun it out because a format can become a standard only if it is not owned by one consultancy's client list.
Business Model
Revenue Model
The local engine and format are free and Apache 2.0. Kerneta says it charges for hosted conversion, cloud hosting/sync, team governance and support rather than for access to the file format; users may also run locally with their own API key, subscription or local model. The company also accepts project supporter contributions.
Pricing Tiers
Self-host the complete local v4.4n engine with an own API key, model subscription or local model; the dashboard also offers three conversions free to try without a key.
$25 of indexing credit; supporter contribution, not a recurring subscription.
$125 of indexing credit and name in the release notes of the next spec version; supporter contribution, not a recurring subscription.
Target Markets
- Developers and AI-agent users
- Startups and enterprises building on MCP or AI assistants
- Teams that need shared, governed AI memory
- Researchers evaluating long-context and memory systems
- Organizations seeking local-first, vendor-independent AI infrastructure
- Long-running AI projects whose history exceeds a model context window
- Persistent memory for coding assistants and agent sessions
- Sharing one memory store across Claude, GPT, Gemini, Cursor, Windsurf and local models
- Reducing repeated context and API/subscription token costs
- Local, private, inspectable and git-able personal or project memory
- Teams needing hosted shared vaults, SSO and audit logs (planned Cloud offering)