Tools — Gnosis Memory

The Difference Is in the Details

Most memory MCPs expose one tool: store_memory(text). Gnosis exposes 14 tools with typed parameters, quality heuristics, and creation guidelines baked into every tool description. The AI reads these descriptions before every call. The protocol teaches structured memory creation without any server-side AI.

Typical memory MCP

store_memory(text: string) Store a memory for later retrieval.

One parameter. No guidance. The AI guesses what to store and how to format it.

Gnosis memory_add

memory_add(content, topics[], macro_topics[], type, ref, artifact, collection) BEFORE generating parameters, reason through: 1. What TYPE? (fact/preference/task/path/decision) 2. What searches should FIND this later? 3. Has this actually happened, or aspirational? CONTENT: First 50 chars = executive summary. Must make sense with zero context.

Seven parameters. Explicit reasoning steps. The AI writes structured, searchable memories because the interface makes it the path of least resistance.

Common Workflows

Session startup: init_core_memories → task_feed
Research: memory_search → memory_retrieve
Task lifecycle: memory_add(type:task) → memory_edit(toggle, output) → memory_edit(set_status:done)
Team handoff: memory_edit(set_status:blocked) → signal_send

init_core_memories

Load Your World in One Call

Run once at session start. Returns your behavioral preferences, topic landscape, active tasks with progress, and pending signals. Previews come back with complete flags so the AI chooses which memories to read in full. Piggyback a search query to save a round trip. Designed to deliver maximum context in minimum tokens.

task_feed — chronological task polling. Up to 50 tasks sorted by last update. Filter by status, topic, or collection. One call covers personal tasks and every shared collection you belong to.

memory_add

Store Knowledge as You Work

Every memory is typed (fact, decision, task, preference, path), tagged with searchable topics, and deduplicated against existing memories. Tasks get markdown checkboxes with a status workflow. The tool description guides the AI through what to store, how to tag it, and how to make it findable later. 99.8% of 2,095 audited memories rated B+ or better.

memory_add_batch — store multiple memories in one call. Same quality rules. Each memory independently typed, tagged, and deduplicated.

memory_search

Find by Meaning, Not Keywords

Semantic search returning up to 32 results. "Database performance" finds query optimization notes that never use the word "performance." Automatically includes all shared collections. Results in under 100ms. Previews first, then the AI retrieves full content only for the 2-3 results it actually needs.

memory_deep_search — broad search returning up to 100 results with cross-encoder reranking. For comprehensive coverage of a topic intersection.

memory_retrieve — get full content by ID after search returns truncated previews.

memory_consolidate — create an executive summary for a topic. Summaries surface near the top of future searches. New sessions get the brief instead of reading 30 individual memories.

memory_edit

Update, Correct, and Operate on Tasks

Three modes: content replace (full rewrite with re-embedding), metadata patch (topics and type only), and task surgery (toggle checkboxes by index, set status, add/remove steps). The output parameter attaches a linked finding to a task — search for the task ID and find both the task and all linked findings together.

memory_delete — permanently delete memories by ID. Single or batch.

collection_manage

Shared Knowledge Across Accounts

Create shared collections, invite by email, set roles, publish personal memories into shared space. Two types: collaborative (everyone reads and writes) and knowledge packs (owner curates, members read). Search automatically includes every collection you belong to. Eight actions: create, list, delete, members, add_user, remove_user, accept, decline, publish.

agent_manage

Isolated Memory Per AI Persona

Create agent identities under your account. Each agent gets its own memory namespace. Set owner_access to control how much of your personal context each agent can see: none (isolated), reader (search your memories), or member (read + write). Add agents to shared collections for team workflows. Four actions: create, list, update, delete.

signal_send

Event-Driven Agent Coordination

Send a signal referencing memory IDs that the recipient should look at. Zero content — just IDs. Recipients see the signal on their next init_core_memories. The recipient must already have collection access to retrieve the referenced memories. Signals auto-expire after 48 hours.

signal_check — check for incoming signals without clearing them.