
A persistent, evolving memory layer for every AI agent
Living Brain is a persistent, self-learning memory layer that gives AI agents shared context and continuity. It compiles scattered conversations, files, and updates into structured, source-linked knowledge, and exposes that context to applications and agents through REST API, SDK, CLI, and MCP. The company reports internal benchmark results on LoCoMo and LongMemEval-S showing high accuracy with drastically reduced token usage and cost compared with full-context baselines.
Automatically turns accumulated knowledge into persistent memory shared across agents and applications, so context doesn't have to be rebuilt each time.
Surfaces non-obvious connections across your documents, conversations, and notes to reveal relationships between pieces of knowledge.
Compiles information into structured, source-linked knowledge, with scoring systems such as Markdown Quality, Memory Worthiness, Placement Fit, Connection Resonance, and Re-entry Priority to organize what is retained and resurfaced.
Compatible with 60+ AI platforms and connectable via REST API, SDK, CLI, and MCP tools (search, pages, index, connections, capture, and create).
Notes and URL capture, document extraction for PDF, DOCX, PPTX, HTML, and text (up to 100 MB per file), audio transcription, automatic chat-turn classification, batch capture of up to 100 items per request, and a Telegram capture bot.
Scheduled daily briefs and page version history (30 days on API, 90 days on Workspace) keep teams up to date on evolving knowledge.
Supports inference-time adaptation via retained knowledge and corrections, and can curate accumulated knowledge, feedback, and task outcomes into training and evaluation datasets for fine-tuning workflows.
Bring documents, conversations, and knowledge into Living Brain via notes, URL capture, file extraction, audio transcription, batch capture, or Telegram.
The compiler and memory policies guide what is retained, organized, and resurfaced, producing structured, source-linked knowledge pages with quality scores.
Semantic search, connection discovery, and brain scores surface relevant context for each task, with median retrieval measured at 464 ms on LoCoMo.
Make the brain's context available to your applications and agents through REST API, SDK, CLI, and MCP, so agents work from accumulated shared memory at inference time.
Pay-as-you-go API plan starts at $5 prepaid with $8 per 1M input tokens and $0.50 per 1M output tokens; Workspace is $99 per month (billed monthly, included at no additional cost with ChatChat Pro+); Business is custom-priced based on deployment scope.
$8 per 1M input tokens + $0.50 per 1M output tokens (prepaid
Pay-as-you-go prepaid memory for applications and agents. No subscription, no monthly minimum. Includes one project, multiple brains, 1 GB storage, 100,000 requests/month, optional auto top-up with monthly cap, and one balance across API, S
$99 per month
Billed monthly; included at no additional cost with ChatChat Pro+. Three team seats, one production project, $80 shared monthly credits for 3 cycles then $50, 10 GB storage, 500,000 requests/month, 90-day page history, 20 webhooks, Slack in
Custom
Tailored to deployment scope. Includes dedicated infrastructure and deployment planning, custom capacity/storage/throughput, data location, retention and access requirements, guided onboarding, private support channel, and custom contracts
Living Brain is a persistent, self-learning memory layer that connects your context, surfaces non-obvious connections, and gives every agent shared memory, context, and continuity. It compiles scattered conversations, files, and updates into structured, source-linked knowledge that can be retrieved via API, SDK, CLI, and MCP.
Yes. You can connect the relevant brain to supported tools and agents via REST API, SDK, CLI, and MCP, helping people work from accumulated knowledge without repeatedly rebuilding context.
In internal evaluations, Living Brain ranked #1 on LoCoMo with 92.2% judged accuracy across 1,540 questions (using ~9.4K context tokens per question vs. ~26K for the full-context baseline). On LongMemEval-S, it showed 0% observed hallucinations under the evaluation's scoring criteria and 94.7% on temporal queries. Note: results are from internal evaluations, not independent audits.
In the internal LoCoMo run, Living Brain achieved 464 ms median memory retrieval (including search and page retrieval) and 3.7 seconds median total answer time (including retrieval and LLM generation).
Living Brain stores knowledge outside the model and retrieves only relevant context for each task, avoiding repeatedly sending entire conversation histories or document collections. In the internal LongMemEval-S evaluation, it used 22.6x fewer tokens per question and achieved 95% lower cost per task versus a 116K-token full-context baseline. Savings depend on workload and configuration.
Yes, in two ways: (1) inference-time adaptation without changing model weights — retained knowledge, corrections, and decisions bring relevant context into future tasks, including for closed-weight models; (2) accumulated knowledge, feedback, and task outcomes can be curated into training examples and evaluation datasets to support fine-tuning or other post-training workflows, where provider training access allows.
Three plans: Living Brain API (pay-as-you-go, $8/1M input tokens + $0.50/1M output tokens, $5 minimum top-up, no subscription), Living Brain Workspace ($99/month, or included with ChatChat Pro+), and Business (custom pricing tailored to deployment scope).
The Living Brain Workspace plan is included with ChatChat Pro+, which covers managing Brains and agents in ChatChat with one credit pool shared by all three team seats.