MindGraph connects your operational data, documents, conversations, and agent activity in one typed memory. Ask what is true now or what was known earlier, trace answers to source evidence, and control what each user or agent can see and do.
Operational data, documents, conversations, and agent activity
Resolves sources into typed, source-linked memory with history and access scope
Shared answers for people, agents, applications, and tools
Powered by MnesticDB · Source-linked · Time-aware · Scoped · Governed
An agent can retrieve one fragment and still miss what it means, when it was true, where it came from, or whether it is allowed to act on it.
MindGraph retrieves relevant source text together with connected objects, claims, and relationships. People can inspect the evidence behind an answer instead of accepting an opaque response.
What did we learn about customer retention from the Q1 report?
Based on the current sources, MindGraph identifies 3 key findings about customer retention:
1. NPS scores improved by 12 points after onboarding redesign
2. Churn rate dropped to 4.2% from 6.8% in Q4
3. User survey shows 89% satisfaction with new features
Hybrid retrieval combines semantically matched passages with connected graph context, then returns both with the answer.
MindGraph is designed around work that spans sources, changes over time, and needs to be explained or controlled.
These examples are generalized from current design-partner conversations. No company information is shown.
Resolve repeated and conflicting requests across a growing conversation corpus. Keep each signal linked to the customer, product, date, and source.
A living product brief shows what customers are asking for, how strongly, what changed, and where more evidence is needed.
Keep the evidence an agent used together with the policy decision, human approval, tool invocation, and resulting artifact.
An operator can verify what the agent knew, why it was allowed to act, who approved it, and what it produced.
Resolve customers, projects, decisions, and open work into one organizational model. Give each person and agent only the context and tools within their scope.
A small team can delegate work without rebuilding company context inside every assistant and application.
MindGraph turns source data into typed objects and source-linked claims, then applies access and policy whenever a person or agent queries or acts.
MindGraph is powered by MnesticDB, our independently maintained hard fork of CozoDB. Postgres is an optional read-only source connector, not MindGraph's database engine.
Identity, access, policy, approval, and execution remain visible from the request through the result.
The credential identifies the person or agent and the organization in which it is operating.
Retrieval is limited to the spaces, projects, and tools granted to that actor.
MindGraph evaluates the proposed action against tool grants and active policies before execution.
A conditional action pauses until an authorized person approves or denies it.
The context, rule, approver, invocation, result, and produced artifact remain linked for review.
MindGraph models the organization, links claims to evidence, preserves change, and governs how people and agents use what it knows.
Turn source schemas, documents, and conversations into reviewed objects and relationships that match how your organization actually works.
Extract claims, goals, and decisions while preserving the exact source material that supports each one.
Ask what was true and what the organization knew at an earlier point without overwriting the history that led there.
Limit knowledge by organization, space, project, role, and identity. Every retrieval uses the effective scope of the person or agent asking.
Grant tools per agent, check policy before execution, pause for approval when required, and keep the decision attached to the result.
People work through chat, briefs, wiki, and the graph explorer. Agents and applications use the same context through SDKs, REST, and MCP.
Self-serve plans include unlimited CRUD, queries, retrieval, and embeddings. Credits are used for ingestion, chat, and wiki compilation. Organization deployments are scoped and priced with you.
Credit costs: document ingestion 2/page, transcripts 4/page, chat 3/message, wiki articles 6/article. Unused credits roll over (up to 2x your monthly allowance).