AI-Native
AI on your data, under the same governance as humans.
AI agents will read your operational data — with or without your permission. Meshbase makes them callers like any other: audit row per call, policies for risky writes, no black box.
Two AI products
MeshIQ for queries. MeshConnect for agents.
Both consume the same federated graph and the same caller-governance pipeline. AI is a first-class citizen, not a separate system.
MeshIQ — natural-language queries
Ask in plain English. MeshIQ translates to a governed graph query, executes under your policy, streams the answer.
Semantic metadata grounding
Each entity and field has a semantic record (description, relationships, query suggestions). NLQ accuracy is grounded in this, not in vibes.
MeshConnect — MCP server
AI agents (Cursor, Claude, your own) connect via MCP under workspace auth. A separate anonymous public-catalog MCP at meshbase.ai/mcp lets agents discover DataProducts, run recommend_plan, and request a signed signup link before any human signs up. Same Meshbase, two scopes: anonymous discovery vs. authenticated execution.
AI = caller, not a special case
Each agent gets a CallerProfile with callerType=AI_AGENT. Same audit, rate limits, mutation policies as a human user.
Mutation policy + approval
Risky mutations route through approval workflows by default for autonomous callers. Reads are open; writes are governed.
AI Act ready
Audit by callerType=AI_AGENT gives compliance reviewers a single answer to 'what did the AI agent do?'.
Why this matters
AI compliance is data compliance.
Without governance, you have a black box. With Meshbase, every AI call leaves an audit row, every risky write goes through approval, and rollback is a query.
Connect a real agent
Bring your agent. See the audit row before you sign.
We connect a Cursor or Claude session to your sandbox workspace and show what AI Act / DPP audit answers look like for your data.