Exa API Returns Empty Results Array Despite Successful Connection

TL;DR

Fix: Collaborator theishangoswami: 'there was an issue with our api, it is now fixed' - it was an Exa-side outage, not a client misconfiguration; reporters confirmed it started working with no changes. Related lesson from the same thread: the npm package can lag the repo (0.3.10 on npm vs 2.0.0 in repo), so update the package before debugging.

The error

The Exa API returns empty results arrays for all queries despite successful HTTP responses (valid requestId, zero cost) through the MCP server on Claude Desktop and Claude Code.

Fix

1. Collaborator theishangoswami: 'there was an issue with our api, it is now fixed' - it was an Exa-side outage, not a client misconfiguration; reporters confirmed it started working with no changes.
2. Related lesson from the same thread: the npm package can lag the repo (0.3.10 on npm vs 2.0.0 in repo), so update the package before debugging.

Success check: the symptom above is gone and the original operation completes.

When to use

• You hit this exact symptom in exa-mcp-server.
• You want the verified fix without digging through the thread.

When not to use

• A different error in exa-mcp-server, even a similar-looking one.
• General exa-mcp-server setup or credential questions.

Compatibility

• Exa MCP Server (exa-labs/exa-mcp-server)

Variant phrasings

• How to fix "exa API Returns Empty Results Array Despite Successful Connection" in exa-mcp-server
• exa-mcp-server: exa API Returns Empty Results Array Despite Successful Connection

Source

• https://github.com/exa-labs/exa-mcp-server/issues/34

In gh:exa-labs/exa-mcp-server#34 (https://github.com/exa-labs/exa-mcp-server/issues/34), the question was: The Exa API returns empty results arrays for all queries despite successful HTTP responses (valid requestId, zero cost) through the MCP server on Claude Desktop and Claude Code. The verified resolution: Collaborator theishangoswami: 'there was an issue with our api, it is now fixed' - it was an Exa-side outage, not a client misconfiguration; reporters confirmed it started working with no changes. Related lesson from the same thread: the npm package can lag the repo (0.3.10 on npm vs 2.0.0 in repo), so update the package before debugging.