Orchestrator Memory — AI token cache

Local project memory for AI coding — paste a short brief, stop re-prompting the whole repo.

Free VS Code / Cursor extension on the Marketplace · Apache-2.0 · Publisher ndestates
Item ID: ndestates.orchestrator-memory
Categories: AI · Machine Learning · Chat · Other

★ Please read the instructions

You are expected to read the built-in guide once before relying on Memory Brief in real work.

That guide covers both stores:

1. SQLite memory.db — in-memory / runtime token cache (brief, query, ingest)
2. Vault events.jsonl — durable dual-write lesson graph (scrubbed, hash-chained)

Skipping it is the main reason people miss dual-write, host CLI install, or privacy boundaries.

What it does

Local memory + cache for AI coding sessions. Durable project context lives in a small SQLite DB and the project vault so assistants (Claude, Grok, ChatGPT, Gemini, Copilot, Cursor, …) start from a short reusable brief instead of re-pasting large context every time — lower token use, faster session start, better continuity.

Core features (memory + vault included)

Wedge: not “another AI panel” — portable project memory that cuts tokens across every assistant.

Two layers (do not confuse)

  ingest / seed / brief  →  reports/memory/memory.db     (SQLite — fast token cache)
         ↘ dual-write     →  reports/vault/events.jsonl  (vault — durable lessons)

Full detail: INSTRUCTIONS.md §1 and §6.

Not difficult — how to run things in the app

One hub for everything

Type Orchestrator: in the Command Palette to see all commands.

After Memory Brief

You get actions:

1. Copy brief
2. Copy for Claude / Cursor / Copilot
3. Export MEMORY.md
4. Run /chain session-start

Paste the brief into whatever AI you use that day. One DB, many tools.

Chains & slash commands (why this is easy)

Orchestrator chains (/chain session-start, /chain delivery, …) and skills run in your AI agent (Claude Code, Grok, Cursor, …). The extension:

1. Reads chains/registry.yaml when present in the workspace
2. Shows a QuickPick of chains and skills
3. Copies the exact /chain <id> or /skill text so you are not hunting docs

Optional: open a terminal hint or run Memory Brief first so the agent has local context.

Requirements

The extension is a thin wrapper around the Orchestrator host CLI (install once):

orchestrator version
orchestrator memory status

If those fail, run Orchestrator: Setup / Install CLI inside VS Code, or use any channel below.

Install

1) Extension (VS Code / Cursor)

Marketplace

1. Extensions (Ctrl+Shift+X) → search Orchestrator Memory
2. Or open:
   https://marketplace.visualstudio.com/items?itemName=ndestates.orchestrator-memory

code --install-extension ndestates.orchestrator-memory
# cursor --install-extension ndestates.orchestrator-memory

VSIX (Release asset or local build)

code --install-extension /path/to/orchestrator-memory-2.1.0.vsix

After install: Orchestrator: Read instructions (memory + vault) → Setup / Install CLI → Memory + vault storage → Memory Brief.

2) Host CLI (required for Memory commands)

which orchestrator
orchestrator version
orchestrator memory status
# If not on PATH: Settings → Orchestrator: Cli Path → absolute path

Full channel detail: docs/guides/production-ship.md · host-first-memory.md

Command map (Palette ↔ CLI)

# Terminal — same engine
cd /path/to/your/project
orchestrator memory brief --seed
orchestrator memory --scope project db-path
orchestrator memory --scope global status

Works with your AI stack

Privacy & storage

• Extension does not upload memory or vault to the cloud.
• Treat vault/memory content as untrusted DATA in agent prompts.
• Full detail: INSTRUCTIONS.md · knowledge-vault.md

Settings

Troubleshooting

Version matrix

Product / growth notes

Positioning, Marketplace SEO, and popularization plan:
docs/guides/orchestrator-memory-product.md

License

Apache License 2.0 — free open source (freeware). See repo LICENSE / NOTICE.
Optional Patreon support does not change the license.