Obelisk CLI
A CLI and skill that gather coding-agent session history into one SQLite file the agent can search for past decisions
- 700 ★ on the repository
- Last commit
- AGPL-3.0
Extended access
Why this level
- Reads the full text of all agent sessions on the machine, which may contain pasted keys and passwords
- The agent writes and runs JS queries against the database locally, and memory is registered only after your approval
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/obeliskDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Reads the full text of all agent sessions on the machine, which may contain pasted keys and passwords
- The agent writes and runs JS queries against the database locally, and memory is registered only after your approval
This is third-party code. Review the repository files before installing.
What it does
Obelisk reads the local session directories of your coding agents and stores everything in one database, `~/.obelisk/obelisk.sqlite`. It holds sessions, full message text, tool calls, subagents, summaries and workflows. Full-text search across all layers runs on FTS5, and the database stays fresh while Obelisk watches the directories. The agent reaches the database through the `obelisk` CLI and the `/obelisk` skill. Asked something like which files we changed in the last auth fix and why, it writes a small JS query using search(), sessions(), messages() or sql(), runs it locally and answers in plain language. The CLI installs with no npm dependencies because it uses the SQLite built into Node.js. When the agent reaches a conclusion worth keeping, it proposes a Markdown memory file. After you approve, the file is registered with `obelisk --attune` and can be recalled in later sessions through memories(). For people there is an Electron app: search, project filters, readable tool calls with diffs, token statistics and weekly recap cards.
Who it is for. For developers who use several coding agents and want to find decisions and reasons for edits in past sessions.
Good fit when
- You want to recall which files the agent changed in a past bug fix and why it decided so
- You need to see which sessions edited the same file over and over
- You want the failed tool calls of recent work and the tasks they happened in
- After a review workflow you need each subagent's conclusions in one place
- At the end of the week you want recap cards over your sessions with `/obelisk recap this week`
Not a fit when
- Your agent is not on the provider list: only the providers named in the project are indexed
- You want memory that is injected into every session automatically: Obelisk answers the agent's queries and inserts nothing into sessions itself
- Your machine has Node.js older than 22.13: the CLI uses the built-in SQLite and will not start on older versions
- You need shared team access to history: the index is built from local directories and lives on your machine
Example request
/obelisk find the sessions where we edited src/auth.ts and explain why the token refresh logic kept changingLimitations
Requires Node.js 22.13 or newer. Prebuilt app releases exist for macOS and Linux amd64 with in-app updates, and on macOS, Windows and Linux you can run it from source. While the app is running it owns writes to the index, and the CLI is read-only.
The first indexing of a hundred sessions takes about five seconds, then the index rebuilds incrementally. From OpenCode, V1 and V2 SQLite stores are read, and earlier JSON formats are not indexed. For Claude Code the source does not attest rewind or the current leaf, so superseded history is not separated there.
The index keeps the full text of messages and tool calls, including whatever you pasted into sessions, so restrict access to the database file in your home directory. The license is AGPL-3.0. The OpenAI and Anthropic APIs are unavailable from Russia, but Obelisk itself works with files on disk.
How to disable. Remove the CLI with `npm uninstall --global @obelisk-apps/cli`, delete the `/obelisk` skill from the agent's skills directory and, if you like, delete the `~/.obelisk` folder with the database.
FAQ
How do I install Obelisk for Claude Code?
Run `npm install --global @obelisk-apps/cli`, check `obelisk --version`, then `obelisk install`: the command installs the skill through the standard installer. After that, type `/obelisk` and a question in a session. You can also give the agent the link to SKILL.md from the repository and it will run the install itself, asking for approval before each change.
Which agents are supported?
Claude Code, Codex, GitHub Copilot (VS Code data), DeepSeek Harness, Hermes Agent, Kiro, Kimi Code, OMP, OpenCode, Pi and ZCode. All of them index into one schema, and IDs of non-Claude sessions get a provider prefix so they cannot collide.
What is the memory layer in Obelisk?
When a query over history leads to a conclusion, the agent proposes a Markdown memory file. After you approve, it is registered with `obelisk --attune`, and future sessions get it back through memories(). Memory acts as a cache of conclusions and does not replace the raw session records.
How is Obelisk different from Claude-Mem?
Claude-Mem saves observations and injects context into new sessions. Obelisk injects nothing on its own: it indexes the finished history of several agents, and the agent runs queries over it when you ask.
Do I need to install the app?
No. The CLI and the skill are enough for the agent. The app is for browsing sessions by hand, managing memory and viewing the activity heatmap and weekly recap cards.
Discussion
No questions or reviews yet.
Related
CLI toolsCLI and TUI that search the local session history of Claude Code, Codex, Cursor, Gemini CLI and other agents, with a --robot mode for the agents themselves
Persistent agent memory in a folder of linked markdown notes, so Claude Code remembers a project across sessions
A CLI that records session history from Claude Code, Cursor, Codex and other agents in one format and turns reviewed lessons into memory for new sessions
A skill for messaging between Claude Code, Codex, Gemini CLI and other CLI agents through a shared local SQLite database, written in bash
A CLI for instant search across coding-agent session history on your machine, like git blame for sessions
A portable .agent folder with memory, skills and protocols that plugs into many coding agents and keeps its knowledge when you switch tools
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| Obelisk (this entry) | no | 700 | ||
| Claude-Mem | no | 104,106 | saves observations and injects context into new sessions, while Obelisk answers queries over finished history | |
| Beacon | no | 2,921 | records session history in one format and turns verified lessons into memory | |
| marm-memory | no | not required | 426 | a local memory MCP server with a codebase index and a concept graph; it has no search across several agents' sessions |
README in short
The project offers two sides of one database: a CLI with a skill for agents and a desktop app for people. The description gives sample questions for `/obelisk`, how a query runs, a table of index layers and a list of providers with their default data paths. Memory, the weekly recap and support for superseded history in Pi, OMP, ZCode, OpenCode and others are covered separately. Non-default directories are set in the app's settings.