MCP Tools

Tools available when connecting to Remind via MCP.

remember

Store an experience as an episode.

remember(content="User prefers TypeScript", episode_type="preference")
remember(content="Use Redis for caching", episode_type="decision", entities="tool:redis,concept:caching")
remember(content="Chose microservices", topic="architecture", source_type="agent")

Parameter	Type	Required	Description
content	string	Yes	The experience to store
episode_type	string	No	observation (default), decision, question, preference, meta, spec, plan, task, outcome, fact
entities	string	No	Comma-separated entity tags (type:name)
topic	string	No	Knowledge area (e.g., "architecture", "product"). Scopes consolidation and retrieval.
source_type	string	No	Origin of the memory (e.g., "agent", "slack", "manual")

ingest

Stream raw text into the auto-ingest pipeline. Text buffers internally until a character threshold (~4000 chars) is reached, then the triage LLM decides what to extract as episodes. A density score (0.0–1.0) may be returned for diagnostics only; it does not gate extraction.

ingest(content="User: Fix the auth bug\nAssistant: Looking at verify_credentials...")
ingest(content="<tool output>", source="tool_output")
ingest(content="Chose Redis for caching", topic="architecture")
ingest(content="<meeting transcript>", instructions="extract decisions and action items")

Parameter	Type	Required	Description
content	string	Yes	Raw text to ingest
source	string	No	Source label for metadata (default: conversation)
topic	string	No	Topic ID or name. When set, all extracted episodes go to this topic. When omitted, the triage LLM infers per-episode topics automatically.
instructions	string	No	Natural-language instructions to steer what the triage LLM extracts (e.g. "focus on architectural decisions"). Appended to the triage system prompt; takes priority over default extraction behavior.

Topic behavior: With topic, all episodes are assigned to the given topic (no inference). Without topic, the triage LLM maps each episode to an existing topic or suggests a new one (auto-created).

Instructions: Use instructions to focus extraction on specific types of information. Without instructions, the triage LLM uses its default behavior (capture anything potentially useful). With instructions, the LLM prioritizes the given directive — useful for ingesting meeting notes, transcripts, or documents where you know what you're looking for.

ingest vs remember: Use remember when you've already decided what's worth storing. Use ingest when you want the triage LLM to choose what to keep — it skips filler and distills memory-worthy episodes. Episodes from ingest are immediately consolidated.

flush_ingest

Force-flush the ingestion buffer, processing whatever text has accumulated regardless of the character threshold.

flush_ingest()
flush_ingest(topic="architecture")
flush_ingest(instructions="extract only action items")

Parameter	Type	Required	Description
topic	string	No	Topic for extracted episodes. Same behavior as ingest() topic parameter.
instructions	string	No	Instructions to steer extraction. Same as ingest() instructions parameter.

Use at session end or whenever you want to ensure all ingested text is processed.

recall

Retrieve relevant memories.

recall(query="authentication issues")
recall(query="auth", entity="file:src/auth.ts")
recall(entity="file:src/auth.ts")
recall(query="database design", topic="architecture")

Parameter	Type	Required	Description
query	string	No	Search query (required for semantic search, not needed for entity-only lookup)
k	integer	No	Number of concepts to return (default: 3)
context	string	No	Additional context to improve retrieval
entity	string	No	Scope to entity (can be used alone without a query)
episode_k	integer	No	Number of episodes to retrieve via direct vector search (default: 5). Set to 0 to disable.
topic	string	No	Filter to a knowledge area. Cross-topic concepts may still surface via spreading activation but are penalized.

At least one of query or entity must be provided.

consolidate

Process episodes into concepts.

consolidate()
consolidate(force=True)

Parameter	Type	Required	Description
force	boolean	No	Force consolidation even with few episodes

inspect

View concepts or episodes.

inspect()                              # List all concepts
inspect(concept_id="abc123")           # Concept details
inspect(show_episodes=True)            # Include episodes

Parameter	Type	Required	Description
concept_id	string	No	Specific concept to inspect
show_episodes	boolean	No	Include episode details

entities

List entities with mention counts.

entities()
entities(entity_type="file", limit=20)

Parameter	Type	Required	Description
entity_type	string	No	Filter by type
limit	integer	No	Max results

inspect_entity

View entity details and relationships.

inspect_entity(entity_id="file:src/auth.ts")
inspect_entity(entity_id="person:alice", show_relations=True)

Parameter	Type	Required	Description
entity_id	string	Yes	Entity ID
show_relations	boolean	No	Include relationships

stats

Memory statistics.

stats()

update_episode

Correct or modify an episode. Updating content resets it for re-consolidation.

update_episode(episode_id="abc123", content="Corrected information")
update_episode(episode_id="abc123", episode_type="decision")

Parameter	Type	Required	Description
episode_id	string	Yes	Episode ID
content	string	No	New content
episode_type	string	No	New type
entities	string	No	New entity tags

delete_episode / restore_episode

Soft delete and restore episodes.

delete_episode(episode_id="abc123")
restore_episode(episode_id="abc123")

update_concept

Refine a concept. Updating summary clears the embedding.

update_concept(concept_id="def456", summary="Refined understanding")
update_concept(concept_id="def456", confidence=0.9, tags="auth,security")

Parameter	Type	Required	Description
concept_id	string	Yes	Concept ID
title	string	No	New title
summary	string	No	New summary
confidence	float	No	New confidence
tags	string	No	Comma-separated tags

delete_concept / restore_concept

Soft delete and restore concepts.

delete_concept(concept_id="def456")
restore_concept(concept_id="def456")

list_deleted

List soft-deleted items.

list_deleted()
list_deleted(item_type="episode")
list_deleted(item_type="concept")

list_topics

List all topics with episode and concept counts.

list_topics()

Returns each topic's name, number of episodes, number of concepts, and when it was last active.

topic_overview

Get top concepts for a specific topic — a quick way to see what Remind knows about a knowledge area without running a query.

topic_overview(topic="architecture")
topic_overview(topic="product", k=10)

Parameter	Type	Required	Description
topic	string	Yes	Topic name
k	integer	No	Number of concepts to return (default: 5)

task_add / task_update_status / list_tasks / list_specs / list_plans

Conditional tools

These tools are only registered when the corresponding episode type (task, spec, plan) is enabled in your configuration. All types are enabled by default.

task_add

Create a new task.

task_add(content="Implement JWT auth", entities="module:auth", priority="p0")
task_add(content="Write tests", depends_on="task-id-1", plan="plan-id")

Parameter	Type	Required	Description
content	string	Yes	Task description
entities	string	No	Comma-separated entities
priority	string	No	p0, p1, p2
plan	string	No	Plan episode ID
spec	string	No	Spec episode ID
depends_on	string	No	Dependency task ID

task_update_status

Transition a task's status.

task_update_status(task_id="abc123", status="in_progress")
task_update_status(task_id="abc123", status="blocked", reason="Waiting on API key")

Parameter	Type	Required	Description
task_id	string	Yes	Task ID
status	string	Yes	todo, in_progress, done, blocked
reason	string	No	Block reason (for blocked status)

list_tasks / list_specs / list_plans

List filtered episodes by type.

list_tasks(status="todo", entity="module:auth")
list_specs()
list_plans()