Codex Supermemory
https://github.com/supermemoryai/codex-supermemory- Category
- AI Agents
- Rank
- No. 1443Tools index
Previous survey · No. 1451 ·
- Pricing
- Open Source
- Type
- TOOL
- Builder
- supermemoryai
- GitHub
- 47 stars
- Latest release
- v1.0.6
- Date
About
Persistent memory for the OpenAI Codex CLI, powered by Supermemory.
What it does
Codex Supermemory connects coding sessions to a hosted memory service through lifecycle hooks. It recalls project context, preferences, decisions, and lessons, then captures conversation material during or after sessions. Repository-derived containers keep memories scoped while allowing clones and linked worktrees to share context.
Why it's ranked here
The design tackles session amnesia with useful operational care: repository identities resist naming collisions, old container formats remain searchable, and hook failures do not interrupt Codex. The current fresh-install behavior is also restrained, favoring session-start profile recall and session-end capture over constant per-prompt activity.
What's good
Installation merges with existing Codex configuration and refuses to write when configuration syntax is invalid. Bundled hook scripts avoid runtime package dependencies. Recall deduplicates facts already shown during a session. Private-tagged content is redacted before upload, while explicit commands provide search, save, forget, status, login, and logout controls.
Tradeoffs
Memory depends on the Supermemory service, authentication, and network access. Privacy redaction requires users to mark sensitive text explicitly. Fresh installs do not recall on every prompt or capture every few turns unless configured. Signal filtering can reduce noisy captures, but the documentation acknowledges that it may miss useful context.
How to use it well
It best suits developers who return to long-running repositories and want accepted decisions, conventions, setup details, and prior fixes carried between Codex sessions. Keep fresh-install defaults for lower activity, enable per-prompt recall only when needed, and mark sensitive passages private. Use explicit commands for important facts. It does not replace source control or maintained project documentation.
Technical notes+
The package is an ESM TypeScript npm CLI defined in package.json, with src/cli.ts installing bundled scripts and skills into Codex. src/config.ts distinguishes fresh installs from legacy upgrades: fresh installs set autoRecallEveryPrompt to false and captureEveryNTurns to zero. src/hooks/recall.ts optionally captures transcript entries and emits additionalContext; src/hooks/flush.ts captures remaining entries on Stop. src/services/tags.ts derives collision-resistant repository tags from normalized Git remotes or local paths, while retaining legacy read tags. src/services/client.ts wraps the Supermemory SDK with 30-second request timeouts, hybrid search, metadata scopes, merged multi-container reads, and error objects. src/services/auth.ts runs a state-token-protected localhost callback and stores credentials with restrictive file permissions.
Observed
- License
- MIT
- Primary language
- TypeScript
- Packaging
- npm package exposing the codex-supermemory executable; published files are limited to dist
- Install surface
- npx CLI with install, uninstall, and status commands
- Interfaces
- Codex lifecycle hooks plus explicit Codex skills for search, add, save, forget, profile, status, login, and logout
- Runtime
- Node.js; hook scripts are pre-bundled with esbuild and described as having no runtime dependencies
- External service
- Uses the Supermemory SDK and defaults to the hosted Supermemory API
- Tests
- The npm test script builds the package and runs a Node unit test file
Read from README.md, package.json, src/cli.ts, src/config.ts, src/version.ts, src/types/index.ts, src/hooks/flush.ts, src/hooks/recall.ts, src/skills/login.ts, src/services/auth.ts, src/services/tags.ts, src/skills/logout.ts, src/skills/status.ts, src/services/client.ts, src/services/logger.ts.
What it can do
Store conversation history persistently
OpenAI Codex CLI interactions → Persistent memory storage
Retrieve past coding conversations
User queries or search terms → Historical Codex interactions
Maintain context across CLI sessions
Previous code generation sessions → Continuous context for new requests
Search through coding history
Keywords or code snippets → Relevant past conversations and code
Remember user coding patterns
Repeated user interactions → Personalized code suggestions
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.