How to Install and Use ByteRover in OpenClaw

ClawHub is OpenClaw's skill registry — it manages the installation, versioning, and configuration of all OpenClaw capabilities. Installing byterover adds structured data exploration to your OpenClaw instance without any manual setup beyond the install command itself. Open your terminal and run the install command below. ClawHub will download the byterover skill package, verify its integrity against the registry checksum, and register it with your local OpenClaw installation. The install process takes about 10-30 seconds depending on your connection speed. After installation completes, you will see a confirmation message listing the skill version and the configuration keys it requires. The skill is installed but not yet active — it will produce a clear error message if you try to use it before setting the API key in the next steps. If you see `clawhub: command not found`, your OpenClaw installation may be incomplete. Reinstall OpenClaw from openclaw.ai to get a fresh ClawHub setup, or check that the OpenClaw binary directory is on your system PATH.

1clawhub install byterover

Navigate to byterover.dev and log in or create an account. Once inside the ByteRover dashboard, click on your account name or avatar in the top-right corner and look for 'API Keys' or 'Developer Settings' in the dropdown menu. Click 'Create New API Key' and give it a descriptive name — something like 'OpenClaw Integration' makes it easy to identify later if you need to revoke it. ByteRover generates the key immediately. Copy it to your clipboard now, as some dashboards only show the full key once at creation time. ByteRover API keys are scoped to your account and all data queries made through OpenClaw count against your ByteRover plan's usage limits. On the free tier, you typically get a limited number of queries per day — check the ByteRover pricing page for current tier details. For heavy use in automated OpenClaw workflows, consider a paid plan that provides higher query limits and priority processing. Store the key somewhere secure like a password manager until you add it to OpenClaw's configuration in the next step. Do not paste it into chat windows or share it publicly.

OpenClaw stores skill credentials in a local configuration directory. The recommended approach is to add your key to the OpenClaw `.env` file or use the `clawhub config set` command — both methods write to the same secure location and keep your key out of shell history. The configuration directory is typically `~/.openclaw/` on macOS and Linux, or `%APPDATA%\openclaw\` on Windows. Open the `.env` file in that directory with any text editor and add your ByteRover API key on its own line. If the `.env` file does not exist yet, create it — OpenClaw will pick it up automatically on next launch. Alternatively, the `clawhub config set` command is the fastest method if you are already in the terminal. It writes directly to the config file without requiring you to open an editor. After saving the configuration, run `clawhub reload` to apply the changes without restarting OpenClaw entirely. Then verify the key is correctly set by running `clawhub config get BYTEROVER_API_KEY` — it should return your key in masked form.

1# Option A: Add to ~/.openclaw/.env
2BYTEROVER_API_KEY=br-your-api-key-here
3
4# Option B: Set via clawhub config command
5clawhub config set BYTEROVER_API_KEY br-your-api-key-here
6
7# Verify the key is saved
8clawhub config get BYTEROVER_API_KEY
9
10# Reload config without full restart
11clawhub reload

With the skill installed and configured, open OpenClaw chat and try a data exploration prompt. ByteRover activates automatically when OpenClaw recognizes data query intent — you do not need to invoke it by name or use any special prefix. Just describe what you want to explore in plain English. For your first test, start with a simple query against a dataset you have already connected to ByteRover. If you have not yet connected a dataset, log into byterover.dev and upload a sample CSV from the dashboard — ByteRover makes connected datasets available to the skill automatically once your API key is configured. Good test prompts are specific: mention the dataset or data type, describe the filter or aggregation you want, and specify the format you want the results in. Vague prompts like 'analyze my data' work but produce less precise results than targeted prompts that describe exactly what insight you need. ByteRover responses include the query it constructed, the results in a structured table or summary format, and a natural language interpretation of what the data shows. If the response does not match what you expected, refine your prompt with more specific terminology or constraints.

ByteRover's real power emerges when you chain multiple analysis steps together in a single OpenClaw conversation. You can start with a broad exploration, then progressively narrow down to specific insights, and finally export or summarize findings — all through natural language turns in the same chat session. For example, you might first ask ByteRover to give you an overview of a sales dataset, then ask it to drill into a specific product category that looks interesting, then filter to a particular time window, and finally summarize the key findings. Each prompt builds on the previous context, so you do not need to re-specify the dataset or filters from scratch. You can also configure default data sources and preferred output formats in the ByteRover skill configuration file. Setting a default dataset means you do not need to mention which dataset to use in every prompt — ByteRover assumes your primary dataset unless you specify otherwise. RapidDev helps teams configure ByteRover alongside other OpenClaw skills for automated data analysis pipelines — particularly useful when combining ByteRover with search skills, document skills, and notification integrations to build end-to-end data workflows.

1# Advanced skill configuration in ~/.openclaw/skills/byterover.yaml
2byterover:
3  default_dataset: "sales_2026"    # default dataset to query (optional)
4  max_rows: 1000                    # max rows returned per query (default: 100)
5  output_format: "table"            # table | json | csv (default: table)
6  include_query: true               # show generated query in response (default: true)
7  timeout_seconds: 30               # max time for a single query (default: 15)