Heptabase CLI

A standalone CLI that wraps the Heptabase MCP server into a command-line tool, built with mcporter.

How It Was Built

Heptabase exposes an MCP server at https://api.heptabase.com/mcp with OAuth authentication. The key insight is using mcp-remote as a stdio adapter — it handles the OAuth browser flow and token caching, then mcporter wraps the resulting stdio MCP server into a CLI.

# 1. Authenticate with Heptabase (opens browser for OAuth)
npx mcp-remote@latest https://api.heptabase.com/mcp --transport http-only

# 2. Generate and compile the CLI
npx mcporter@latest generate-cli \
  --command 'npx -y mcp-remote@latest https://api.heptabase.com/mcp --transport http-only' \
  --output heptabase-cli.ts \
  --compile heptabase \
  --description "Heptabase knowledge base CLI"

Authentication

• First run opens a browser for Heptabase OAuth login
• Tokens are cached in ~/.mcp-auth/ and auto-refresh
• To force re-login: rm -rf ~/.mcp-auth/

Installation

As an Agent Skill

Install as a skill for Claude Code, Cursor, Codex, and other agents:

npx skills add madeyexz/heptabase-cli

This makes /heptabase-cli available as a slash command in your agent.

Via bunx (no install needed)

Requires Bun.

bunx heptabase-cli search-whiteboards --keywords "project"

Download binary from GitHub Releases

curl -L https://github.com/madeyexz/heptabase-cli/releases/latest/download/heptabase -o heptabase
chmod +x heptabase
sudo mv heptabase /usr/local/bin/

Build from source

Requires Node.js and Bun.

# 1. Authenticate with Heptabase (opens browser for OAuth)
npx mcp-remote@latest https://api.heptabase.com/mcp --transport http-only

# 2. Generate and compile
npx mcporter@latest generate-cli \
  --command 'npx -y mcp-remote@latest https://api.heptabase.com/mcp --transport http-only' \
  --output ./heptabase-cli.ts \
  --compile ./heptabase \
  --description "Heptabase knowledge base CLI"

# 3. Add to PATH
sudo ln -sf "$(pwd)/heptabase" /usr/local/bin/heptabase

Verify

heptabase --help

Usage

# Search
heptabase search-whiteboards --keywords "topic1,topic2"
heptabase semantic-search-objects --queries "machine learning" --result-object-types card

# Read
heptabase get-object --object-id <id> --object-type card
heptabase get-whiteboard-with-objects --whiteboard-id <id>
heptabase get-journal-range --start-date 2026-01-01 --end-date 2026-02-21

# Write
heptabase save-to-note-card --content "# Title\n\nBody text"
heptabase append-to-journal --content "Some entry"

# PDF
heptabase search-pdf-content --pdf-card-id <id> --keywords "term1,term2"
heptabase get-pdf-pages --pdf-card-id <id> --start-page-number 1 --end-page-number 5

# Output formats: text (default), json, markdown, raw
heptabase search-whiteboards --keywords "project" --output json

For MCP setup with Claude Code, Cursor, ChatGPT, etc., see the official Heptabase MCP docs.

Available Commands

| Command | Description | |---|---| | search-whiteboards | Search whiteboards by keywords | | semantic-search-objects | Hybrid full-text + semantic search across cards, journals, PDFs, highlights | | get-object | Get full content of a card, journal, media, highlight, etc. | | get-whiteboard-with-objects | Get all objects and connections on a whiteboard | | get-journal-range | Fetch journal entries for a date range (max 92 days per call) | | save-to-note-card | Create a new note card in your Inbox | | append-to-journal | Append content to today's journal | | search-pdf-content | BM25 keyword search within a PDF (up to 80 ranked chunks) | | get-pdf-pages | Get specific page ranges from a PDF card |

Project Structure

heptabase-cli/
├── heptabase          # Compiled standalone binary (bun)
├── heptabase-cli.ts   # Generated TypeScript source
├── package.json       # npm package config (for bunx heptabase-cli)
├── SKILL.md           # Agent skill definition (skills.sh)
├── config/
│   └── mcporter.json  # mcporter server configuration
└── README.md