Skip to content

Coralogix docs

Use the Coralogix MCP server to search and read the official Coralogix documentation portal (coralogix.com/docs) from your AI coding assistant. Two read-only tools — search_docs and fetch_doc — let Cursor, Claude Code, Codex, or any other MCP client ground answers in the real docs alongside your telemetry queries, with no extra setup.

The same lookup is also available in the Coralogix CLI as cx docs search and cx docs fetch.

What you need

  • MCP server v2 — send the mcp-version: v2 header (see Setup).
  • No Coralogix API key or OAuth token. The documentation portal is public, so these two tools sit alongside the telemetry tools without going through authentication.

Tools

ToolDescription
search_docsKeyword search across the Coralogix docs portal. Returns ranked title and path-suffix matches.
fetch_docRetrieve the Markdown body of one doc page, given a path suffix from search_docs.

Both tools read the published llms.txt index and the per-page index.md files the docs portal serves. There is no HTML scraping — your agent reads the exact Markdown the docs team publishes.

Search the docs portal

Use search_docs to find pages by keyword.
ParameterRequiredDescription
queryYesShort keywords. Examples: "explore spans", "OpenTelemetry traces", "API keys".
limitNoMaximum number of matches to return, 1 to 20. Defaults to 5.

The response is a numbered list of ranked matches, each as <rank>. <title> — <suffix>:

1. Explore spans — user-guides/data_exploration/spans
2. Spans tab — user-guides/apm/spans-tab
3. Trace explorer — user-guides/apm/trace-explorer

The suffix on each match is the path under coralogix.com/docs/ and is what fetch_doc consumes. If no result clears the relevance threshold, search_docs returns a "no docs matched" error instead of low-confidence matches.

Retrieve a doc page

Use fetch_doc to get the Markdown body of a page returned by search_docs.
ParameterRequiredDescription
suffixYesPath suffix returned by search_docs (for example, user-guides/data_exploration/spans). Leading slashes, trailing slashes, and a docs/ prefix are all tolerated. Do not pass a full URL — use a suffix.

The response is the doc page's path plus its full Markdown body:

Path: user-guides/data_exploration/spans

# Explore spans

The MCP server caches the index for one hour, so consecutive searches do not download llms.txt again.

Standard workflow

  1. search_docs with two to four focused keywords.
  2. Pick the best suffix from the ranked matches.
  3. fetch_doc on that suffix to read the page.
  4. Run search_docs again with different wording, or fetch_doc an adjacent page, if the first match was incomplete.

When to use these tools versus other tools

Use docs tools whenUse telemetry tools when
You need to know how a Coralogix feature works or how to configure it.You need to query your own logs, spans, metrics, or alerts.
You want to cite an official documentation page in an answer.You already have a precise DataPrime or PromQL query.
You are setting up an integration, agent, or SDK.You want to look up incidents or alert events by ID.

For tenant data, use the Logs and traces, Metrics, Real User Monitoring, or Alerts and incidents tools instead.

Troubleshooting

SymptomCauseResolution
search_docs returns "no docs matched".Query did not reach the relevance threshold.Broaden or rephrase. Try the feature name plus a category — for example, "OpenTelemetry traces" instead of "distributed tracing setup".
fetch_doc rejects the input as a URL.Caller passed a full URL instead of a suffix.Pass the suffix from search_docs (path under /docs/), not the rendered page URL.
fetch_doc returns an HTTP error.The page moved or was removed.Run search_docs again and use a fresh suffix from the ranked matches.
Search results look outdated.The MCP server caches the docs index for one hour.Wait for the cache window to roll, then retry.

Next steps

To run the same lookup from a terminal, install the Coralogix CLI and use cx docs search <query> and cx docs fetch <suffix>.