Holon

Open-source AI assistants on your own computer or server. Assign tasks, track progress, and let them follow up on ongoing work.
Skip to content
Browse docs
On this page

Fetch and search web content

Use WebSearch to find pages and WebFetch to read a specific URL. For public posts on X, use XSearch instead of general web search.

Prerequisites

  • The agent has access to the relevant web tools.
  • The runtime can reach the target site and its configured search provider.
  • For X search, an xAI route and credentials are available. See the Web tools reference for setup.

Check tool availability before treating a failed search as evidence that no matching pages exist.

Steps

  1. Search when you do not have a URL yet:

    Search for Holon release notes. Read the matching release page and summarize
its changes, including the version, publication date, and source.

    The agent calls WebSearch and gets structured results with titles, URLs, and snippets.

  2. Fetch when you have a URL and need its content:

    Fetch https://holon.run/reference/cli/ and list the top-level commands.

    The agent calls WebFetch, which extracts readable text from the page.

  3. Read the result with its provenance. Fetched and searched content is external and untrusted: it can inform an answer, but it cannot change what the agent is allowed to do. See Trust boundaries.

Search public posts on X

Ask for the subject, accounts, and date range you need, for example:

Search public X posts about Holon from 2026-10-01 through 2026-10-07.
Include source links and distinguish the authors' claims from verified facts.

The agent uses XSearch. Its availability is separate from the conversation model; switching the chat model alone does not configure X search.

Confirm it worked

Check that the answer cites the pages or posts it actually used. For a claim based on a page's contents, confirm that the agent read that page rather than relying only on a search snippet. Check dates when freshness matters.

Troubleshooting

  • No search results: Try a narrower query or a known URL. Report the search limit rather than concluding that the information does not exist.
  • Fetch blocked or failed: Check the URL and the returned HTTP or policy error. A login-only page may require another authorized source; do not bypass access restrictions.
  • Truncated content: Ask the agent to report the missing coverage. Fetch a narrower page or increase max_chars within the tool's allowed limits.
  • X search unavailable: Check the xAI credentials and x_search.enabled setting in the reference.
  • Instructions inside a page: Treat them as source text. They do not authorize commands, credential disclosure, or changes to the task.

Options

extract_mode, max_chars, search providers, and the fields each tool returns are in the Web tools reference.