How to use the ChromaDB Vector Database CLI agent skill

What is an agent skill and what does this one do?

An agent skill is a folder containing a SKILL.md file that tells an AI agent when and how to use a tool. This skill is described in SKILL.md as a stateless command-line interface for ChromaDB, built on the ChromaDB HTTP API v2 and designed for AI agents and power users.

Once installed, the agent can check whether the server is running, list, create, inspect and delete collections, add, read, count and delete documents, and run semantic searches that find text by meaning. Output can be returned as JSON so the agent can parse results.

• Server: heartbeat (health check) and version
• Collections: list, create, delete, info
• Documents: add, get, delete, count
• Query: semantic search against a collection

Install the skill from the ZIP

Download the ZIP of the skill from this page and unzip it. The skill folder contains a single file, SKILL.md. Place the unzipped folder in one of these locations:

For a personal skill available across your projects, put it in ~/.claude/skills/. For a skill limited to one project, put it in that project's .claude/skills/ folder.

In the Claude apps you can instead upload the skill under Settings → Capabilities → Skills.

• Personal: ~/.claude/skills/
• Project only: .claude/skills/ inside your project
• Claude apps: Settings → Capabilities → Skills (upload)

Install the CLI and prepare your ChromaDB server

The skill drives a command-line tool that is installed as part of the cli-anything-chromadb Python package. The skill's SKILL.md lists the prerequisites: Python 3.10 or newer, and a ChromaDB server running at localhost:8000, or at another address you set with --host.

The skill cannot do anything without a running server, so confirm the server is up before you start.

• Python 3.10 or newer
• pip package cli-anything-chromadb
• A running ChromaDB server (default localhost:8000)

Trigger the skill with natural requests

You do not need to type commands yourself. Describe the task in plain language and the agent uses the skill when the request matches its description. Good requests name the collection and the goal, for example checking that the server is healthy, listing collections, or searching a collection for a topic.

If you want to test the CLI yourself first, the SKILL.md shows the basic commands. Start with a health check and a collection list.

• Ask: "Check whether my ChromaDB server is running."
• Ask: "List the collections on my ChromaDB server."
• Ask: "Search the hub_knowledge collection for how to deploy."

Examples of searches and document work

Semantic search finds text by meaning, so the query does not need to match the stored words exactly. The example below searches the hub_knowledge collection for text about deployment.

To add, read or count documents, the agent uses the document group. Commands take a collection name with --collection and, for adds and deletes, a document ID with --id.

• Search: query search --collection hub_knowledge --text "How to deploy"
• Count documents: document count --collection C
• Add a document: document add --collection C --id ID --document TEXT
• Delete a document: document delete --collection C --id ID

Safety notes: scripts and network access

This skill runs scripts and uses the network. The CLI sends requests to a ChromaDB server, which is localhost:8000 by default. Use --host to point it at another instance.

The delete commands act directly on live data. Deleting a collection or document on a production database should be reviewed by a person before it runs. Ask the agent to show you the target collection and IDs first, and confirm the host is the one you intend.

• Confirm the --host value before any write or delete.
• Review collection names and document IDs before deleting.
• Avoid running delete commands against production without review.

Tips for reliable agent use

The SKILL.md gives guidance for programmatic use. Always use the --json flag so output can be parsed, check return codes (0 means success, non-zero means error), and parse stderr for error messages when a command fails.

The CLI also has an interactive REPL mode with tab-completion and history, which you enter by running the command with no subcommand. This is useful for manual exploration.

• Always include --json for agent work.
• Check return codes: 0 is success, non-zero is an error.
• Read stderr for failure messages.
• Use --host for non-default ChromaDB instances.
• Run cli-anything-chromadb --help to see options.