Lithtrix — Identity, Memory & Trust for AI Agents

Identity, memory, trust, and swarm primitives for AI agents. MIRC base always free.

Lithtrix is an MCP server that supplies foundational primitives for AI agent infrastructure: identity, memory, trust, and swarm coordination. It is published by ai.lithtrix under the package name lithtrix-mcp on npm at version 0.20.2 and communicates over stdio transport. Configuration requires a single environment variable, LITHTRIX_API_KEY, which must be set before the server can be used.

The server is described as targeting multi-agent systems that need consistent identity management, persistent or shared memory across agent sessions, trust mechanisms between agents or between agents and external systems, and swarm-level coordination primitives. The base MIRC tier is stated to be permanently free, making it accessible for initial development and evaluation without upfront cost.

Because no tool list is exposed in this record, the specific operations an agent can invoke at runtime are not documented here. Developers should consult the repository at github.com/lithtrix/lithtrix-mcp for the current API surface and any usage limits that apply beyond the free MIRC base. The server is not a database, does not provide SQL query access, and is not a fit for workloads that require only straightforward relational data retrieval without agent-to-agent coordination concerns.

Use cases

  • Assign and manage persistent identities for individual AI agents across sessions
  • Share or persist memory state across multiple agents in a swarm
  • Establish trust relationships between agents or between an agent and an external service
  • Coordinate behavior across a group of agents using swarm primitives
  • Bootstrap a multi-agent project using the free MIRC base tier before scaling

When to use it

  • Building multi-agent systems that require shared identity or memory across agent instances
  • Implementing trust or authorization logic between cooperating AI agents
  • Prototyping swarm-style agent architectures without immediate infrastructure cost
  • Integrating via stdio transport with an MCP-compatible host environment

When not to use it

  • The workload only requires direct SQL or relational database access with no agent coordination
  • The runtime cannot supply the required LITHTRIX_API_KEY environment variable
  • The host expects a non-stdio MCP transport such as HTTP SSE
  • A full tool manifest is needed before adoption, since no tools are currently enumerated in the published record
  • The project has no need for identity, memory, or trust abstractions and simply needs data retrieval