frinus-mcp MCP Server

φρήν + νοῦς — Gives your Claude agent long-term, cognitive memory backed by the Frinus platform.

Local serverstdioMIT

What is the frinus-mcp Server?

frinus-mcp is an MCP project maintained in the frinus-ai/frinus-mcp repository. The maintainer describes its purpose as “φρήν + νοῦς — Gives your Claude agent long-term, cognitive memory backed by the Frinus platform”. Read the project documentation for the complete scope.

Installation

The documented npm package is frinus-mcp. The configuration below comes from the maintainer’s example; adapt its paths and account values to your environment.

Configuration values

The example declares FRINUS_API_KEY. Secret-looking values have been replaced with placeholders. Required and optional labels are shown only when the documentation states them.

Source

Project documentation · Repository · Discovery listing. This profile records the public documentation checked on 2026-10-08.

Recorded tools

Check the linked publisher documentation and your connected client's tool list. Names and availability can vary with the server version and account permissions.

ToolWhat it does
memory_storeCreate a memory. agent_id, content required. memory_type ∈ {episodic, semantic, procedural}. scope ∈ {user, agent, universe, organization}. importance 0–1.
memory_searchSemantic similarity search. Filters: agent_id, memory_types, limit.
memory_getFetch a memory by memory_id (full content, metadata, relevance).
memory_listList memories for an agent_id with optional type filter.
memory_deletePermanently delete a memory by memory_id.
memory_reinforceBoost relevance of a useful memory (memory_id, boost 0–1, default 0.1).
memory_weakenPenalise an outdated memory (memory_id, penalty 0–1, default 0.2).
working_memory_getLoad current state for a context. Always call at task start.
working_memory_addPersist current state. Context formats: agent:{uuid}, universe:{uuid}, organization:{uuid}.
working_memory_clearRemove all items for a context.
session_startBegin a session for an agent. Returns session_id. Supports parent_session_id for subagent inheritance.
session_endTerminate a session and finalise its summary.
session_contextCombined working + long-term retrieval, enhanced with extracted session topics.
session_summaryGenerate a structured recap (decisions, learnings, pending items).
session_clearWipe a session's working state without ending it.
stream_captureRecord an input / output / internal note tied to a session.
stream_get_sessionReplay every captured item for a session_id.
stream_get_recentRecent stream items across sessions (filterable).
stream_processPromote pending items to long-term memory (manual trigger; scheduler also runs every 5 min).
memory_get_contextBuild a token-bounded context window for a task description.
search_with_attentionRAG with task-type-aware weighting. task_type ∈ {implementation, debug, deploy, documentation, review} drives memory_type weights.
user_registerRegister a user with the memory system (idempotent).
user_get_contextRetrieve combined user memories and tenant context.
heartbeat_tickCheap tick for an agent — drives relevance decay and lightweight consolidation.
sleep_runTrigger a sleep cycle. phases ⊆ {evaluation, forgetting, consolidation, relevance}. Mirrors the scheduler's normal/deep sleeps.
consolidation_detect_conflictsSurface candidate conflicting memories above a similarity threshold.
consolidation_resolve_conflictKeep one memory, supersede the other with a written resolution note.
consolidation_detect_redundantFind near-duplicates ready for merge or removal.
hierarchy_consolidateRoll several related memories into a higher-level summary memory.
hierarchy_get_treeInspect the consolidation tree for a root memory.
agent_createCreate an agent (optionally from a template) scoped to a universe / team.
agent_listList agents in the caller's tenant. Auto-scoped via API key.
agent_getFetch an agent by agent_id.
agent_updateUpdate persona, team, universe, etc.
agent_deleteDelete an agent.
agent_invokeProgrammatically invoke an agent with a task. Returns its tool calls and final answer. Used by delegation.
universe_createCreate a universe in the caller's org. Slug + name + description.
universe_listList universes for the resolved org.
universe_updatePatch a universe (name, description).
universe_hierarchyWalk the full L0 → L3 tree for a universe with status icons.
training_teachInject a fact or procedure. type ∈ {semantic, procedural}.
training_qaTrain with pairs of {question, answer}.
training_statsCoverage statistics across the corpus.
training_gapsDetected gaps in knowledge / weakly-covered topics.
training_recentMost recently ingested memories from training.
task_createCreate a task (title, description, optional assigned_agent_id, parent task).
task_getFetch a task with its full state.
task_listList tasks with filters (status, agent, parent).
task_updateUpdate status, output, or assignment.
skill_listEnumerate available skills in the tenant.
skill_assignAttach a skill to an agent.
skill_removeDetach a skill from an agent.
skill_agent_listList the skills owned by a given agent.
credential_storeStore an encrypted credential under a ref (e.g. mysql_x).
credential_getInspect non-secret metadata only (host/user/db/port + the env vars credential_exec will inject). Never returns the value.
credential_execRun a command with the credential injected into the child process ENV. Returns only stdout/stderr/exit_code — the secret never reaches the model, screen, or…
credential_listList stored credential refs (no secret data).
credential_deleteDelete a stored credential.

How to install the frinus-mcp Server

{
  "mcpServers": {
    "frinus": {
      "command": "npx",
      "args": [
        "-y",
        "frinus-mcp@latest"
      ],
      "env": {
        "FRINUS_API_KEY": "YOUR_FRINUS_API_KEY"
      }
    }
  }
}

From the project documentation. Replace placeholders and example paths before connecting.

Configuration

VariableDescriptionRequired
FRINUS_API_KEYDeclared in the maintainer’s client configuration.Yes