Ori Memory

Persistent memory infrastructure for AI agents.

Connector by aayoawoyemi · GitHub account verified by the MCP Registry

IndexedApache-2.0Local (stdio)★ 333

Install

Works with Claude Code, Codex, VS Code, Cursor, Windsurf, Claude Desktop. Read the repository before you give any tool access to your data.

npx agentmarket install connector/aayoawoyemi-ori-memory

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

About

Persistent memory infrastructure for AI agents. Identity, knowledge graph, and search.

Use cases: agent-memory, knowledge-base

Reviews

Write a review

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this server can do and what it asks for, from its tool list and install config. Checked Oct 25, 2026.

Tools: 6 read · 5 write · 2 delete · 3 not yet classified
Secrets: None asked for
Install scripts: None found
Runs as: Node.js package

Tools by what they do 16

Read from the source code; the running server may list different tools.

Read 6 · Look things up without changing anything.

memory_sql
ori_explore
ori_health
ori_query_ranked
ori_query_similar
ori_validate

Write 5 · Create, change, send or run things.

ori_add
ori_index_build
ori_update
ori_update_decision
ori_whats_new

Delete 2 · Delete, drop, revoke or wipe things.

ori_learning_reset
ori_prune

Not yet classified 3 · We could not tell from the name and description.

ori_promote
ori_wake
ori_warmth

Files it touches

Any path the agent passes to its tools

Tools (16)

Read from the source code; the running server may differ. Descriptions are the server’s own words; hidden characters are shown as [U+…].

memory_sql
Read-only SQL over your memory index (SQLite). Use for questions the ranking tools cannot express. Views: v_note (slug, title, type, modified, access_count, inbound, outbound, pagerank, betweenness, q_value, exposure_count), v_link (src, dst edges), v_dangling (missing link targets and who cites them), v_retrieval (every note ever returned to you: session_id, timestamp, query_text, slug, rank, final_score), v_session, v_stage. Also note_project(note_id, project) and co_occurrence. Single SELECT/WITH/EXPLAIN/VALUES only, strings in single quotes, 200 rows and 2s max. Pass schema=true for full D…
ori_add
Create a note in inbox. Always pass description and project: you know them now, and a note without them fails the vault schema.
ori_explore
Deep memory exploration via PPR graph traversal. Propagates through wiki-links at exploration-tuned parameters (α=0.45), returns ranked notes with content snippets. Use for multi-hop questions, connection discovery, and deep recall. Heavier than ori_query_ranked but finds notes flat retrieval misses.
ori_health
Full diagnostic
ori_index_build
Build or update the embedding index. Only re-embeds changed notes unless force=true. Also bootstraps co-occurrence edges from wiki-link structure.
ori_learning_reset
Record the user's answer to Ori's learning-reset question (learning_reset_notice). Call ONLY after the user explicitly answers. decision=accepted clears learned ranking scores (notes are untouched; the database is backed up first). decision=declined keeps them and permanently stops the question. If the user says 'not now', do not call this; the question is raised at most 3 times in total.
ori_promote
Promote an inbox note to notes/ with classification, linking, and area assignment. YOU are the intelligence layer — read the note, decide its type, write a description, identify links to existing notes, and pass your decisions as overrides. Heuristics run as fallback for anything you don't specify.
ori_prune
Analyze activation topology and identify archive candidates. Dry-run by default. Set apply=true to archive.
ori_query_ranked
Full ranked retrieval with Q-value reranking, co-occurrence PPR, and stage meta-learning. 4 base signals (composite + keyword + graph + warmth) fused via RRF, then Phase B Q-value reranking. Excludes archived notes by default. Triggers spreading activation.
ori_query_similar
Composite vector search only (semantic + metadata, no keyword/graph). Faster but single-signal. Excludes archived notes by default.
ori_update
Update agent files: identity, goals, methodology (self/), or daily, reminders (ops/). Auto-backs up previous version before writing.
ori_update_decision
Record the user's answer to an Ori update notice. Call ONLY after the user explicitly answers. decision=accepted after a successful update; decision=declined if they want to stay. If the user defers, do not call this — the reminder resurfaces next session.
ori_validate
Validate a note against schema
ori_wake
Session boot: constant-size briefing, budget-capped. Call this first in a session. On a vault with no identity written yet it also returns an onboarding script.
ori_warmth
Associative warmth field for the current context. Returns low-token note titles, scores, and sources showing what memory is resonating before and alongside retrieval.

Show 1 more tools

ori_whats_new
Release notes for an Ori version (default: the installed one). Use when the user asks what changed, or to read notes cut short in a whats_new notice.

Signals

Stars: 333
Forks: 31
Open issues: 1
Last commit: Oct 1, 2026 (28 days ago)
Last release: Oct 1, 2026 (28 days ago)

From GitHub (fetched Oct 24, 2026).

Details

Registry name: io.github.aayoawoyemi/ori-memory
Version: 0.3.0
Auth: No auth
Packages: npm ori-memory
License: Apache-2.0
Repository: github.com/aayoawoyemi/Ori-Mnemos
First seen: Oct 24, 2026
Last checked: Oct 25, 2026
Source: Official MCP Registry github-graphql

Cite this page

Ori Memory MCP server. Agent Mart. https://agentmart.dev/connectors/aayoawoyemi-ori-memory (last updated 2026-10-25)