Signet CLI
Signet keeps a shared memory for AI agents: memories, session transcripts and secrets, with every fact linked to its source
- 314 ★ on the repository
- Last commit
Broad access
Why this level
- Stores secrets and session transcripts, and the background service reads data from Obsidian, GitHub, Notion and Discord
- Writes hooks and plugins into the settings of Claude Code, Codex and other agents and mixes its memory into prompts
- Installation can run as a script from the internet that downloads a binary and runs with the user's rights
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/signet-ai-signetaiDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Stores secrets and session transcripts, and the background service reads data from Obsidian, GitHub, Notion and Discord
- Writes hooks and plugins into the settings of Claude Code, Codex and other agents and mixes its memory into prompts
- Installation can run as a script from the internet that downloads a binary and runs with the user's rights
This is third-party code. Review the repository files before installing.
What it does
Signet is a memory service for AI agents that runs in the background on your machine or as a server for a team. It builds memories from session transcripts, imported files and connected sources (Obsidian, GitHub, Notion, Discord, web pages) and hands the agent what is relevant before the next prompt. Every fact links to its source fragment, so you can trace it back to the original. A process called dreaming reads new data together with old, links people, projects, facts and relationships into a knowledge graph, revisits contradictions and proposes updates with citations. Signet connects to agents through their hooks, plugins or extensions: hooks and MCP for Claude Code, a native plugin for Codex, and separate integrations for OpenCode, OpenClaw, Hermes Agent, Gemini CLI and others. Agents send each other messages that survive restarts and arrive at the start of the recipient's next session. The CLI runs it: `signet setup` prepares a workspace, `signet status` checks the service and `signet dashboard` opens a panel with the memory graph and a chat over memory. Secrets are stored encrypted.
Who it is for. Developers who work with several agents at once, such as Claude Code and Codex, and do not want to explain the project again after switching tools or models.
Good fit when
- You switch from Claude Code to Codex or back and project agreements must stay
- Context sits in Obsidian, GitHub issues and Notion pages and the agent needs a digest of it before a prompt
- You need to find out why the agent believes a fact: Signet shows history, competing claims and exact source passages
- The team runs a shared memory server and agents need separate read permissions
- Project secrets should be stored encrypted and handed to agents as needed
Not a fit when
- A plain AGENTS.md or CLAUDE.md with project rules is enough: a separate memory service with a background process is excessive
- Company policy forbids any outgoing traffic from installed tools, even anonymous
- You only need a simple notes log without a knowledge graph or background processing
- Sources such as Slack, email, Telegram and WhatsApp are critical: the README marks them as coming soon
Example request
Why do you believe we chose PostgreSQL for this project? Show where that came fromLimitations
- Signet sends anonymous telemetry: install and version counts, feature usage, cost totals per provider and sanitized crash reports. Memory content and queries are not sent. It is switched off with the `telemetryEnabled: false` setting or the `SIGNET_TELEMETRY_OPTOUT=1` variable.
- Nightly builds (`--nightly`) carry unreleased work and can break, and the stable channel is recommended for daily use.
- Document import supports Word, PowerPoint, Excel, OpenDocument, RTF, EPUB, CSV and PDF, and no other formats are listed.
- Slack, email, Telegram and WhatsApp as sources are marked as coming soon.
How to disable. Disable Signet in the settings of each agent where it is wired in through hooks or a plugin. Full removal is described in the documentation at docs.signetai.sh, and the npm package is removed with `npm uninstall -g signetai`.
FAQ
How do I install Signet for Claude Code?
Run `npm install -g signetai`, then `signet setup`: the wizard picks a provider and connects sources and agents. Claude Code gets hooks and MCP. Check the state with `signet status`.
What is dreaming?
It is background processing of memory. Signet reads fresh data together with what is already stored, links people, projects and facts into a graph, revisits contradictions and proposes updates. All changes are validated and recorded with citations.
Which agents are supported?
Claude Code, Codex and the ChatGPT desktop app, OpenCode, OpenClaw, Hermes Agent, Kimi Code, Pi, Oh My Pi, Gemini CLI, ForgeCode and Muse Code. Each connects its own way: hooks and MCP, a plugin or an extension.
Where are secrets stored?
Encrypted. The master key sits in the system keyring, and without one secrets go to an encrypted file and the dashboard shows a warning. Keep the keychain in your recovery plan.
What does the telemetry send?
Install and version counts, feature usage, cost totals per provider and sanitized crash reports. Memory content, prompts and search queries are not sent, and every event is also written to a local log in the workspace.
Discussion
No questions or reviews yet.
Related
CLI toolsSibyl is a self-hostable knowledge graph shared by Claude Code, Codex and Cursor for decisions, gotchas and conventions
agentcairn keeps Claude Code, Codex and Cursor memory in Markdown files for Obsidian and hands an agent only what it needs
Shared memory for AI agents across machines: a fact one agent writes shows up for the others with no server in between
Encephalon stores project decisions and conventions as JSON inside Git so coding agents can search them between tasks.
Persistent agent memory in a folder of linked markdown notes, so Claude Code remembers a project across sessions
Shared local memory for AI agents: Claude Code, Codex, Cursor and others find the agreements of past sessions in it and record new ones
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| Signet (this card) | no | 314 | ||
| agent-memory | no | 4,315 | memory on Markdown files with a SQLite index, while Signet adds a knowledge graph, sources and dreaming background processing | |
| Supermemory | yes | OAuth | 31,602 | a remote memory MCP server with plugins and an API, while Signet runs locally or on your server and stores secrets |
README in short
The README opens with a warning that Signet has no token or coin. It then covers a quick start, how it works, tables of sources and formats, the list of agent harnesses, the dashboard and desktop app, trust inspection of memory, telemetry and LongMemEval benchmark results.