Ori Mnemos

Open-source persistent memory infrastructure for AI agents.

Ori implements human cognition as mathematical models on a knowledge graph. Activation decay from ACT-R. Spreading activation along wiki-link edges. Hebbian co-occurrence from retrieval patterns. Reinforcement learning on retrieval itself. Recursive graph traversal with sub-question decomposition. The system learns what matters, forgets what doesn't, and optimizes its own retrieval pipeline.

Persistent memory across sessions, clients, and machines. Zero-infrastructure retrieval that matches and in several cases strongly outperforms incumbents on benchmarks — and you own every byte of your data. Markdown on disk. Wiki-links as graph edges. Git as version control. No database lock-in, no cloud dependency, no vendor capture.

v0.7.0 · npm · Paper · Apache-2.0

Use

Ori is three surfaces over one index. The markdown is the truth; the index is derived. The learned half — Q-values, LinUCB arms, retrieval history — is not, so export it before deleting anything (see When to rebuild).

CLI

npx ori init          # scaffold a vault
npx ori index build   # derive the index
npx ori explore "…"   # navigated retrieval
npx ori sql "…"       # read-only SQL over the index

MCP server — ori serve, registered in a client config. This is how an agent uses it.

Library — recall is the same wired entry the CLI and the MCP ori_recall tool both go through, so the programmatic path and the agent path cannot drift.

import { recall } from "ori-memory";

const res = await recall("./vault", "what did we decide about caching?", { limit: 5 });
for (const hit of res.data.results) console.log(hit.title, hit.score);

searchComposite is also exported for callers that have already assembled vectors, graph metrics and a config; recall does that assembly for you.

The export surface is deliberately small and is a semver contract; the rest of src/core is internal. Versions before 0.8.0 shipped no main and no exports, so a bare import threw and the library path did not exist — but the CLI and MCP paths always worked, and existing users were unaffected.

ori-memory/cli resolves to the CLI entry, for callers that need to locate the binary and spawn it rather than link against it. require.resolve on it is the intended use; importing it runs the CLI.

Benchmarks

ForgetEval — Can It Forget On Command?

deeplethe/lethe, bench/forgeteval/, MIT. No API key, no network, no LLM judge. The scorer is a ~20-line deterministic substring check in GeneratedCase.run(): it calls recall_texts(query, k=10) itself, joins the top 10, and tests must_contain / must_not_contain. Generation is random.Random(42) over templates. The optional LLM hook is llm=None by default and was not used.

Not fitted to the suite: unseen seeds give 97.2% (seed 7) and 98.0% (seed 123). The fixes were structural bugs in the matcher, not case-specific patches.

Quote these against a 35% floor, not against zero. The oracle is must_contain AND must_not_contain over the top 10, and when must_contain is empty — every decay case and 150 of 200 purge cases, 350 of 1,000 — a system that returns nothing passes vacuously. A null adapter that accepts writes and never returns anything scores 350/1000 = 35%, including 100% on the whole decay family. On the 650 cases that actually discriminate, Ori is 628/650 = 96.6% and its purge drops from 91% to 64%, which is its real weak spot. The published LangMem 99.5 / Lethe 99.3 / Mem0 88.8 are full-suite and carry the same floor. See docs/falsification/forgeteval-validity.md.

Ori scored 0/1000 on this benchmark earlier the same day. Not a low score — a structural zero. ForgetEval's adapter protocol has three optional operations, supersede, release and purge, and Ori had none of them: zero source hits across src/. Every case was N/A. The ACT-R decay and Ebbinghaus curves Ori already had are ranking-time priors, and no benchmark measures those; ForgetEval's "decay" family means an explicit release(query) call. src/core/forget.ts is 305 lines and closed the whole gap in a day, which is the most informative number on this page.

That table is not a ranking, and "third" would be a bad way to read it.

ForgetEval's code lives inside deeplethe/lethe — the benchmark and its top-scoring system are the same org, in a repo with 14 stars. The template column has four entries, one of which (MemPalace) scores 0 by construction because it exposes no deletion primitive at all. Two of the rest saturate. On the adversarial layer Ori is 5th of 14 configurations and lands inside the 63–68% band the paper's own McNemar test calls noise (χ²=0.125, p=0.724); the paper's words are "the bench reads the trade-off, not a winner." The one comparison that is statistically real is Ori vs Lethe on template, z=2.81, p=0.005 — Lethe is genuinely ahead.

Who is missing matters more than who placed. Supermemory (30.6k stars, $2.6M seed) ships POST /v4/memories/forget-matching — natural-language forgetting with dryRun, threshold, maxForget and an audit handle — plus versioned PATCH supersession. That is a better-specified control plane than anything scored here, and it maps onto the adapter protocol almost verbatim. It has never been benchmarked. Neither have Hindsight (24.0k), Cognee (30.8k, excluded for API incompatibility), MemOS (11.5k) or Honcho (7.3k).

So the honest claim is not that Ori forgets better than the field. It is that Ori forgets offline, with no API key, and that the field has no idea how well it forgets:

Fifteen agent-memory systems were checked. Zero publish a forgetting benchmark for their own system. Five such benchmarks exist — ForgetEval, Memora/FAMA, MemoryAgentBench-SF, StateMemBench, MemLeak — and vendors cite none of them. Every forgetting number in existence was produced by a rival or an outsider.

Read the adversarial column with one more caveat. 253 of its 385 cases were admitted only if the vendor's own system passed them, annotated in adversarial.py as "Oracle-validated (Lethe / Lethe+LLM passes the case)". The authors' own blind 77-case external subset drops the whole field from the 63–68% band to 28–33%, which says the in-house suite is materially easier. A benchmark whose admission filter is "the measurer's system solves it" cannot rank the measurer.

Reproduce:

git clone https://github.com/deeplethe/lethe && cd lethe
cp <ori>/bench/forgeteval_ori_adapter.py bench/forgeteval/ori_adapter.py
export ORI_BRIDGE=<ori>/bench/forgeteval-bridge.mjs
python -m bench.forgeteval.run --adapter ori --suite template --scale 200 --seed 42

Four minutes, 1,000 cases, $0.00. The adapter talks NDJSON to a resident Node process because the harness makes ~10 calls per case and a CLI subprocess per call would spend hours on interpreter startup.

HotpotQA — Multi-Hop Retrieval

Head-to-head against Mem0. Both systems indexed the same documents and answered the same questions in the same run.

n = 50, single run, no seed averaging, topK = 5. Mem0 at 1.0.6 (March 2026); 2.x is not yet re-run, so read this as a point-in-time comparison, not a current one. Raw output: bench/results/, reproduce with bench/hotpotqa-eval.ts and bench/mem0-hotpotqa.py.

These are not HotpotQA's official metrics and must not be compared to the HotpotQA leaderboard. The official scorer, hotpot_evaluate_v1.py, reports answer EM/F1 under its own normalize_answer, plus supporting-fact F1 over (title, sentence_id) pairs, plus joint EM/F1. The table above is a title-level retrieval metric defined in bench/hotpotqa-eval.ts, and "answer proxy" is not a HotpotQA metric at all — it is token recall of the gold answer against retrieved text. The comparison is valid in one direction only: Ori and Mem0 went through the same harness on the same questions, so the ratio between the two columns means something. The absolute numbers do not transfer anywhere.

At n = 50 the Wilson 95% intervals are Ori [0.75, 0.94] and Mem0 [0.18, 0.43] on Recall@5. They do not overlap, so the gap is real, but the two-decimal precision in the table is not: read 0.87 as "high 0.80s".

Latency is not reported here. The evaluation harness does not record it, so any number would be recalled rather than measured. What is measured is that Ori answers from markdown plus a local SQLite index with no API key and no network.

LoCoMo — Long-Term Conversational Memory

1,536 questions over 10 conversations. Retrieval is BM25 + embedding + PageRank fusion at top-5. No API key and no network: the answer column is an extractive proxy, token recall of the ground-truth answer against retrieved text, not a generated answer.

Raw output: bench/results/locomo-eval-2026-09-19T22-37-31-698Z.json. Reproduce with npx tsx bench/locomo-eval.ts --json; the run takes 48 s.

Multi-hop and temporal are the weak categories and are reported as such.

Two corrections to earlier versions of this file, both found on 2026-09-19:

• It previously reported 695 questions and an overall recall of 0.687. That

subset silently excluded the open-domain category, which is 841 of the 1,536 questions — more than half the benchmark. The per-category figures were close to correct; the "overall" was an average over a hand-picked three categories. bench/README.md carried a third set of numbers again (44.7% recall) that reproduces nothing in the current harness. One number now, with the run file beside it.

• The 2026-09-19 run reproduces the 2026-07-22 run to three decimals, so these

figures are stable across the retrieval-metric repair in 09ac45d and the lambda change in 72fdd13.

No comparison table against published LoCoMo leaderboards is given, on purpose. Those are LLM-judge scores; the above is token F1. They are different quantities and putting them in one column would invent a ranking rather than report one. A previous version of this README did exactly that.

LoCoMo itself has known defects. An independent audit found 6.4% of questions carry wrong answer keys, putting the theoretical ceiling at 93.57%, and the standard gpt-4o-mini judge accepts 62.81% of deliberately wrong answers. At least one published score exceeds the mathematical ceiling. A close result on this benchmark is weak evidence in either direction, which is why it is reported here and not led with.

LongMemEval-S — Retrieval, 500 Questions

Session granularity, 470 scored (the official scorer excludes the 30 abstention questions). No API key, no network, no LLM judge — the benchmark's own scorer imports sys, json and numpy and nothing else.

recall_all@k requires every gold session in the top k; recall_any@k requires one. The official summary reports recall_all@5 and ndcg_any@5.

Against published figures on the same benchmark using the same embedder, all three zero-API-call:

Ori leads at @1 and is at parity by @10. That is a smaller claim than it looks. An independent analysis of MemPalace (arXiv:2604.21284) concluded its 96.6% R@5 "is the performance of ChromaDB's default embedding model (all-MiniLM-L6-v2) applied to verbatim text chunks" and is reproducible with a minimal ChromaDB setup. At k=5 this metric is saturated and mostly measures the embedder, which is the same one in all three rows. The honest reading is that Ori's retrieval is not the bottleneck and this axis no longer separates systems.

Ori's recall_any@5 of 0.9660 and MemPalace's published 96.6% agree to three significant figures. That is a coincidence, not a copied number; the full per-question output is committed.

Weak categories, consistent with LoCoMo: multi-session 0.653 and temporal-reasoning 0.772 recall_all@5, against 1.000 for both single-session types. Multi-hop and temporal are where Ori loses on both benchmarks, which is two independent measurements agreeing rather than noise.

Reproduce:

npx tsx bench/longmemeval-eval.ts --data <longmemeval_s_cleaned.json>
python bench/longmemeval-score.py <rankings.json> <path/to/LongMemEval>
python <LongMemEval>/src/evaluation/print_retrieval_metrics.py <rankings.jsonl>

11 minutes, 500 questions, $0.00. bench/longmemeval-score.py imports the benchmark's own evaluate_retrieval rather than reimplementing recall_all@k, so these are the authors' metric definitions.

Quick Start

npm install -g ori-memory
ori init my-agent
cd my-agent

Connect to your agent:

# Full adapters — auto-orient at session start, capture at session end
ori bridge claude-code --vault ~/brain                 # hooks + MCP + CLAUDE.md
ori bridge hermes --vault ~/brain                      # native plugin + MCP + HERMES.md
ori bridge opencode --vault ~/brain                    # plugin + MCP + AGENTS.md

# MCP-only adapters — tools available, no lifecycle automation
ori bridge cursor --vault ~/brain                      # .cursor/mcp.json
ori bridge codex --vault ~/brain                       # ~/.codex/config.toml

# Any MCP client
ori bridge generic --vault ~/brain                     # prints config for manual setup

Claude Code, Hermes Agent, and OpenCode get full lifecycle integration — the agent orients at session start, captures insights at session end, and validates notes on write. Cursor, Codex, and other MCP clients get access to all 16 tools but manage their own session lifecycle.

Manual MCP config (works with any client that speaks MCP):

{
  "mcpServers": {
    "ori": {
      "command": "ori",
      "args": ["serve", "--mcp", "--vault", "/path/to/brain"],
      "env": { "ORI_VAULT": "/path/to/brain" }
    }
  }
}

Start a session. The agent receives its identity automatically and begins onboarding on first run.

What's New

v0.6.0 — Navigated Recursion. ori explore no longer returns a flat synthesis. The agent sees the decomposition tree — which branches produced results, which hit dead ends — and steers the traversal itself. New session commands: explore-start, explore-expand, explore-conclude. Budget is a nudge, not a wall: soft exhaustion with explicit extension. A cross-encoder reranking stage now sits on top of four-signal fusion. RMH Constraint 2 goes from partial to real.

$ ori explore-start "why did we choose SQLite over postgres"

exploration e7f2 — 3 branches
├─ [1] storage engine tradeoffs        4 notes, strong signal
├─ [2] deployment constraints          2 notes
└─ [3] prior migration decisions       dead end — no notes

next: ori explore-expand e7f2 1   |   ori explore-conclude e7f2 --answered

v0.5.6 — OpenCode bridge. Full lifecycle integration: first-run onboarding, auto session capture, note validation, multi-vault support. ori bridge opencode — one command.

v0.5.5 — Ebbinghaus warmth. Notes accessed once fade fast (half-life ~7 days). Notes accessed across many sessions embed deeply (up to ~28 days). Short-term and long-term memory, structurally distinct.

Full history in the CHANGELOG.

Recursive Memory Harness

Ori is the first implementation of the Recursive Memory Harness (RMH) framework — a set of constraints on how persistent memory should behave for AI agents.

The core insight comes from Recursive Language Models (Zhang, Krassa & Khattab, 2026). RLM treats context not as input to be stuffed into a window, but as an environment to be navigated. The model doesn't get a bigger desk — it gets legs and walks into the library. RMH applies the same principle to persistent memory.

Three constraints define the framework:

1. Retrieval must follow the graph. Memory is not a flat vector store. Notes are nodes, wiki-links are edges. Retrieval walks the structure — Personalized PageRank at α=0.45, spreading activation along edges, community-aware traversal. The topology of the graph shapes what gets found.

1. Unresolved queries must recurse. When a single retrieval pass is insufficient, the system decomposes the question into sub-questions, retrieves against each, and synthesizes. Convergence detection stops recursion when new passes stop surfacing new information. This is what ori explore does.

1. Every retrieval must reshape the graph. Retrieval is not read-only. Co-occurrence edges grow between notes retrieved together (Hebbian learning). Q-values update based on whether retrieved notes were actually useful. The graph learns from how it is used — every query makes the next query better.

Most memory systems treat retrieval as search. RMH treats retrieval as navigation, recursion, and learning — on a graph that evolves with every session.

Read the full paper: Introducing Recursive Memory Harness

What It Does

• Persistent identity. Agent state — name, personality, goals, methodology — is stored in plain markdown and auto-injected at session start via MCP instructions. Identity survives client switches, machine migrations, and model changes without reconfiguration.

• Knowledge graph. Every [[wiki-link]] is a directed edge. PageRank authority, Louvain community detection, betweenness centrality, bridge detection, orphan and dangling link analysis. Structure is queryable through MCP tools and CLI.

• Three memory spaces. Identity (self/) decays at 0.1x — barely fades. Knowledge (notes/) decays at 1.0x — lives and dies by relevance. Operations (ops/) decays at 3.0x — burns hot and clears itself. The separation is architectural, not cosmetic.

• Cognitive forgetting. Notes decay using ACT-R base-level learning equations, not arbitrary TTLs. Used notes stay alive. Their neighbors stay warm through spreading activation along wiki-link edges. Structurally critical nodes are protected by Tarjan's algorithm. ori prune analyzes the full activation topology before archiving anything.

• Four-signal fusion. Semantic embeddings, BM25 keyword matching, personalized PageRank, and associative warmth fused through score-weighted Reciprocal Rank Fusion. Intent classification (episodic, procedural, semantic, decision) shifts signal weights automatically.

• Dampening pipeline. Three post-fusion stages validated by ablation testing: gravity dampening halves cosine-similarity ghosts with zero query-term overlap, hub dampening applies a P90 degree penalty to prevent map notes from dominating results, and resolution boost surfaces actionable knowledge (decisions, learnings) over passive observation.

• Learning retrieval (v0.4.0). Three intelligence layers improve retrieval quality from session to session, synthesized from 63 research sources. See Retrieval Intelligence below.

• Capture-promote pipeline. ori add captures to inbox. ori promote classifies (idea, decision, learning, insight, blocker, opportunity), detects links, suggests areas. 50+ heuristic patterns. Optional LLM enhancement.

• Zero cloud dependencies. Local embeddings via all-MiniLM-L6-v2 running in-process. SQLite for vectors and intelligence state. Everything on your filesystem. Zero API keys required for core functionality.

Retrieval Intelligence (v0.4.0)

Three learning layers that improve retrieval quality over time without manual tuning. Synthesized from 63 research sources across reinforcement learning, information retrieval, cognitive science, and bandit theory.

Layer 1 — Q-Value Reranking

Notes earn Q-values from session outcomes via exponential moving average updates. Over time, genuinely useful notes rise and noise sinks.

After RRF fusion, Phase B reranks the candidate set with a lambda blend of similarity score and learned Q-value, plus an exploration bonus that ensures under-retrieved notes still get discovered. The bonus shrinks as a note is shown without ever being used; once a note has been credited it keeps the full bonus, so being used never ranks a note below an equally relevant note nobody has seen. A cumulative bias cap (MAX=3.0, compression=0.3) prevents runaway score inflation.

Layer 2 — Co-Occurrence Edges

Notes that are retrieved together grow edges between them — Hebbian learning on the knowledge graph. Edge weights are computed using NPMI normalization (genuine association beyond base rate), GloVe power-law frequency scaling, and Ebbinghaus decay with strength accumulation (frequently co-retrieved pairs decay slower).

Per-node Turrigiano homeostasis prevents hub notes from absorbing all edge weight. Bibliographic coupling bootstraps day-0 edges from existing wiki-link structure before any queries have been run.

The combined wiki-link + co-occurrence graph feeds a Personalized PageRank walk (HippoRAG, α=0.5) that surfaces notes semantic search alone would never find.

Layer 3 — Stage Meta-Learning

Each optional pipeline stage (PageRank, warmth, hub dampening, co-occurrence PPR) is wrapped in a LinUCB contextual bandit with an 8-dimensional query feature vector. The system learns which stages help for which query types and auto-skips stages that consistently hurt.

Three-way decisions per stage: run / skip / abstain (stop the pipeline early). Cost-sensitive thresholds ensure expensive stages face a higher bar. Essential stages (semantic search, BM25, RRF fusion, Q-reranking) never skip. Q-reranking is essential because the bandit's reward measures result shape and term recall, which cannot see whether a note was used. An ACQO two-phase curriculum runs all stages during exploration (first 50 samples), then optimizes.

Session Learning Loop

Query → Retrieve → Use (cite, update, create) → Reward signals
  ↓                                                    ↓
  Co-occurrence edges grow                Q-values update (session-end batch)
  ↓                                                    ↓
  Stage meta-learner updates              Better retrieval next session

All updates happen in a single SQLite transaction at session end, in order: co-occurrence → Q-values → stage learning.

The Stack

Layer 6: MCP Server                    16 tools, 5 resources — any agent talks to this
Layer 5: Recursive Exploration         PPR graph traversal, sub-question decomposition, convergence detection
Layer 4: Retrieval Intelligence        Q-value reranking, co-occurrence learning, stage meta-optimization
Layer 3: Dampening Pipeline            gravity, hub, resolution — ablation-validated
Layer 2: Four-Signal Fusion            semantic + BM25 + PageRank + warmth → score-weighted RRF
Layer 1: Knowledge Graph + Vitality    wiki-links, ACT-R decay, spreading activation, zone classification
Layer 0: Markdown files on disk        git-friendly, human-readable, portable

16 MCP tools · 5 resources · 20 CLI commands · 874 tests

Token Economics

Without retrieval, every question requires dumping the entire vault into context. With Ori, the cost stays flat.

A typical session costs ~$0.10 with Ori. Without it: ~$6.00+.

Architecture

                          Any MCP Client
                    (Claude, Cursor, Windsurf,
                     Cline, Hermes, custom agents, VPS)
                              │
                        MCP Protocol
                        (stdio / JSON-RPC)
                              │
                    ┌───────────────────┐
                    │    Ori MCP Server  │
                    │                   │
                    │  instructions     │   identity auto-injected
                    │  resources  (5)   │   ori:// endpoints
                    │  tools    (16)    │   full memory operations
                    └─────────┬─────────┘
                              │
            ┌─────────────────┼─────────────────┐
            │                 │                 │
      ┌───────────┐    ┌───────────┐    ┌───────────┐
      │ Knowledge │    │ Identity  │    │Operations │
      │   Graph   │    │  Layer    │    │  Layer    │
      │           │    │           │    │           │
      │  notes/   │    │  self/    │    │  ops/     │
      │  inbox/   │    │  identity │    │  daily    │
      │  templates│    │  goals    │    │  reminders│
      └─────┬─────┘    │  method.  │    │  sessions │
            │          └───────────┘    └───────────┘
      ┌─────┴──────┐
      │            │
   Wiki-link   Embedding        ┌──────────────────────┐
    Graph       Index            │ Retrieval Intelligence│
   (in-mem)    (SQLite)          │                      │
      │            │             │  Q-values  (note_q)  │
   PageRank    Semantic          │  Co-occur  (edges)   │
   Spreading   BM25              │  Stage Q   (LinUCB)  │
   Activation  4-Signal          │  Dampening (3 stages)│
   Communities Fusion            │  Explore   (PPR+RMH) │
                                 └──────────────────────┘

MCP Tools

Five tools were removed in favour of memory_sql, which answers their questions in one statement and is 20–46× faster because it reads the index instead of re-walking the vault from disk:

The three navigated-exploration tools (explore_start / _expand / _conclude) were removed from MCP after six months with zero recorded sessions. They remain available as ori explore-start, ori explore-expand and ori explore-conclude.

CLI

# Vault management
ori init [dir]                    # Scaffold a new vault
ori status                        # Vault overview
ori health                        # Full diagnostics
ori whats-new [version]           # Release notes

# Note lifecycle
ori add <title> [--type <type>]   # Capture to inbox
ori promote [note] [--all]        # Promote to knowledge graph
ori validate <path>               # Schema validation
ori archive [--dry-run]           # Archive stale notes
ori prune [--apply] [--verbose]   # Topology analysis + archive candidates

# Retrieval
ori explore <query>               # Recursive graph traversal (RMH)
ori query ranked <query>          # Full intelligent retrieval
ori query similar <query>         # Semantic search
ori query important               # PageRank ranking
ori query fading                  # Vitality detection
ori query orphans                 # Notes with no incoming links
ori query dangling                # Broken wiki-links
ori query backlinks <note>        # What links to this note
ori query cross-project           # Multi-project notes

# Infrastructure
ori index build [--force]         # Rebuild everything derived from the vault
ori index status                  # Index statistics
ori graph metrics                 # PageRank, centrality
ori graph communities             # Louvain clustering
ori serve --mcp [--vault <path>]                                # Run MCP server
ori bridge claude-code [--scope <s>] [--activation <a>] [--vault <p>]  # Claude Code (hooks + MCP + instructions)
ori bridge hermes [--scope <s>] [--activation <a>] [--vault <p>]       # Hermes Agent (plugin + MCP + instructions)
ori bridge opencode [--scope <s>] [--activation <a>] [--vault <p>]     # OpenCode (plugin + MCP + AGENTS.md)
ori bridge cursor [--scope <s>] [--vault <p>]                          # Cursor (MCP config)
ori bridge codex [--scope <s>] [--vault <p>]                           # Codex (TOML config)
ori bridge generic [--scope <s>] [--vault <p>] [--json]                # Any MCP client (prints config)
ori bridge status [--json]                                             # Inspect all bridge installs
ori bridge <target> --uninstall                                        # Remove Ori config for a target

Path-taking commands treat relative file paths as vault-relative. Absolute paths continue to work unchanged.

When to rebuild

.ori/ holds two different things, and only one of them is disposable.

Derived — rebuilt from the markdown on demand:

Accumulated — nowhere else, and a rebuild destroys it:

retrieval_log, note_q, q_history, q_history_genuine, stage_q, stage_log, boosts, note_access, memory_events, and the retrieval rows of co_occurrence.

Measured on a 1,548-note vault: rm -rf .ori/ && ori index build empties eight of those tables — six months of retrieval history across 564 sessions, 1,427 learned Q-values, and the eight live LinUCB arms that decide which ranking stages run. Two of them, q_history_genuine and memory_events, are created by no production code at all and cannot be reconstructed by anything.

So export before you delete:

ori index export-learned          # -> ops/ori-learned.ndjson
rm -rf .ori/
ori index build
ori index import-learned

That round-trip is verified lossless on a real vault by bench/learned-roundtrip.mjs, which compares row counts and a digest over the Q-values themselves, because counts matching while values drift is a pass that means nothing. The export is NDJSON, deterministically ordered, ~23 MB for 56,850 rows against a 243 MB binary — small enough and diffable enough to commit, which the binary is not.

The practical rule: edit and query freely; run `ori index build` after bulk-adding notes from outside Ori, and `ori index build --force` if a query reports it cannot cover the vault. --force is a full reparse that keeps accumulated state; only deleting the file loses it.

Earlier versions of this README said everything under .ori/ was derived and disposable. That was wrong, and following it cost six months of learning.

Vault Structure

vault/
├── .ori                       # Vault marker
├── ori.config.yaml            # Configuration
├── notes/                     # Knowledge graph (flat, no subfolders)
│   └── index.md               # Hub entry point
├── inbox/                     # Capture buffer
├── templates/                 # Note and map schemas
├── self/                      # Agent identity
│   ├── identity.md            # Name, personality, values
│   ├── goals.md               # Active threads, priorities
│   ├── methodology.md         # Processing principles
│   └── memory/                # Agent's accumulated insights
└── ops/                       # Operational state
    ├── daily.md               # Today's completed and pending
    ├── reminders.md           # Time-bound commitments
    └── sessions/              # Session logs

Every file is plain markdown. Open it in any text editor, Obsidian, or your file browser. git log is your audit trail.

Deployment

Local. Install globally, ori init, connect your MCP client. Done.

VPS / headless. Install on the server. ori serve --mcp --vault /path/to/vault. Memory persists on the filesystem. Back up with git push.

Remote terminals. Hermes Agent supports Docker, SSH, Modal, and Daytona backends. If your agent runs in a remote terminal, ori must be installed and on PATH inside that environment, and the vault must be on persistent storage (not ephemeral). For serverless backends like Modal where environments hibernate, mount the vault on a persistent volume.

Multi-vault. Separate Ori instances for separate agents. Each vault is self-contained: its own identity, knowledge graph, and operational state.

Scriptable. CLI returns structured JSON. Use in cron jobs, webhook handlers, or orchestration loops.

Install Model

Ori separates three install concepts:

• scope: global follows one vault across the machine, project stays inside one repo/workspace
• activation: auto runs ori_wake at session start where the adapter supports it, manual leaves tools available but does not auto-orient
• vault: explicit --vault wins; otherwise Ori resolves by install scope

Precedence rules:

• project install overrides global install
• explicit --vault overrides inferred vault
• project activation overrides global activation

Bridge lifecycle:

• rerun the same ori bridge ... command to update vault path or activation in place
• use --uninstall to remove Ori-owned config from supported adapters
• generic installs emit manual uninstall instructions because Ori does not own that client config surface

Claude Code, Hermes Agent, and OpenCode are fully automated adapters with lifecycle hooks. Claude Code uses hook scripts; Hermes uses a native Python plugin installed at ~/.hermes/plugins/ori/; OpenCode uses a JavaScript plugin at .opencode/plugins/lifecycle.js. All three auto-orient at session start (via first-run detection) and capture insights at session idle. Cursor and Codex have native MCP config install support. Codex writes to ~/.codex/config.toml and uses a single global config surface; "project" scope there means project-like runtime vault discovery, not a separate project config file. Other MCP-capable clients can use ori bridge generic now and wire the emitted config into their own client surface.

Configuration

ori.config.yaml controls all tunable parameters. Generated with sensible defaults on ori init.

LLM integration is optional. Every operation works deterministically with heuristics alone. When configured, LLM improves classification and link suggestions.

Why Sovereignty Matters

Most memory systems store your agent's knowledge in infrastructure you do not control. A proprietary database. A cloud service. A vendor's format.

Ori stores memory as files you own. The vault is portable. Move it to a new machine, push it to a git remote, open it in a text editor. Switch MCP clients by changing one config line. The memory survives any platform change because it was never locked to a platform.

This is not ideological. It is architectural. Portable memory is composable memory.

Development

git clone https://github.com/aayoawoyemi/Ori-Mnemos.git
cd Ori-Mnemos
npm install
npm run build
npm link
ori --version

npm test              # 860+ tests
npm run lint          # Type check
npm run dev           # Watch mode

See CONTRIBUTING.md before opening a PR.

Thanks to @maichler and the rest of the Ori community for their PRs and additions.

License

Apache-2.0

Memory is sovereignty. Ori gives your agent a mind.