Skip to content

Getting Started

One command. No cloning, no venv, no config files.

Terminal window
claude mcp add -s user hive -- uvx --upgrade hive-vault

The --upgrade flag ensures you always get the latest version from PyPI on each session start. uvx handles the Python environment automatically.

The uvx configurations above resolve the current release for each new session. If you use a persistent installation for daemon mode, upgrade it deliberately:

Terminal window
# Linux / macOS
uv tool upgrade hive-vault
# Windows (the version is optional; omitted selects the latest PyPI release)
hive self-upgrade [version]

On Windows, self-upgrade builds the new release beside the active files and atomically switches the managed runtime, so it does not modify files held by a running Hive process. Open a new terminal after the first managed upgrade so the launcher added to your user PATH is available. Start a new MCP session after any upgrade; an existing session keeps its current server process.

Once registered, your AI assistant can use Hive tools. Try asking:

“Use vault_list to see my vault”

The assistant will call vault_list() and list all projects with file counts and available shortcuts.

By default, Hive looks for your vault at ~/Projects/knowledge. If this path doesn’t exist, vault tools will return a helpful error with setup instructions — the server still starts, so worker tools remain available.

To point Hive to your vault:

Terminal window
claude mcp add -s user hive -e VAULT_PATH=/path/to/your/vault -- uvx --upgrade hive-vault

To enable task delegation to cheaper models:

Terminal window
claude mcp add -s user hive \
-e VAULT_PATH=/path/to/your/vault \
-e HIVE_OLLAMA_ENDPOINT=http://your-ollama:11434 \
-e OPENROUTER_API_KEY=sk-or-... \
-- uvx --upgrade hive-vault

See Configuration for all environment variables.

Run a health check:

“Use vault_health to check my vault”

You should see project counts, file counts, and staleness metrics for each project in your vault.