IndraDBCapability / 01
Persistent reasoning
Record decisions and the reasons for them. Keep that record for the next session.
Memory operations / local first
Memory for AI agents.
Keep a record beyond the session. Indra lets your agent save its reasoning, search past decisions, and explore other paths.
01 / The problem
A new session can lose the reason.
Without a saved record, an agent cannot refer to yesterday’s decision. It can repeat work or suggest a different approach. Indra provides tools to save that context and find it again.
02 / Capabilities
What Indra gives your agent
Capability / 02
Branching exploration
Try another approach on a separate branch. Keep the main line of work intact.
Capability / 03
Semantic search
Search past decisions by meaning. Local embeddings are on by default.
Capability / 04
Reasoning history
Read the record of changes. Compare branches or points in history.
Capability / 05
3D visualization
View saved entries as points in the web app. Explore groups of related entries.
03 / Operation
How it works
The MCP server connects your agent to the Indra database. The agent calls memory tools as part of its work.
Make a decision
You ask for advice. The agent considers the context.
Record the reason
The agent calls
indra_rememberwith its decision and reason.Find it next session
The agent calls
indra_searchto find past reasoning. This record is a tool for consistent decisions.
Mission log / example, not a recorded result
User: Should I use PostgreSQL or MongoDB?
indra_search({ query: "database recommendations" })
Found: “Recommended PostgreSQL for projects with relational data...”
Agent: Based on your e-commerce use case and my previous analysis, PostgreSQL is the better choice. You need ACID transactions for orders.
indra_remember({ content: "Recommended PostgreSQL for e-commerce. Relational catalog + transaction needs.", id: "ecom-db" })