MCP100 Assessment

Editorial verdict

Recommended for bounded web and code research when query privacy, result verification, service cost, and downstream use of retrieved content are actively managed.

Reviewed July 24, 2026

Task Fit

Mapped

Specialized workflows

Trust

Reviewed

A dated editorial review is published on the MCP detail page.

Maintenance

Activity observed

A dated repository or package update is available; editorial maintenance review is still pending.

Popularity

Signals available

Visible usage signals are shown for discovery only, not as proof of safety.

Best for

Search the web, retrieve pages, find code, and run research workflows through Exa.
Teams already using Exa that can test with non-sensitive data and least privilege.

Not ideal for

Users who need an MCP100-recommended recommendation today
Unattended production workflows before permissions and failure behavior are reviewed

Avoid if

The connected data or account cannot be safely exposed to the selected AI client
You cannot verify least-privilege access and review write-capable tool calls

Trade-offs

First-party provenance and a clear task improve confidence, but do not replace an independent permissions and usability review.
Client transport compatibility does not prove that authentication, scopes, failure handling, and every tool behave safely.

Evidence scope

Evidence

Current source snapshot

Popularity

5k MCP repository stars

248k package downloads / 30 days

Repository identity: exa-labs/exa-mcp-server

Package identity: exa-mcp-server · version 3.4.1

Stars observed September 24, 2026 · Downloads observed September 24, 2026

Exact reviewed repository identity; current GitHub repository stars are eligible as repository-level adoption evidence.

Exact MCP-specific package identity verified against the reviewed canonical repository.

Maintenance

Publisher activity observed September 23, 2026

Current source snapshot

Trust review

Source relationship: First-party MCP

Reviewed July 24, 2026

Sources

GitHub repository ↗Package registry ↗

Evidence-backed setup

Install & Agents

Choose your MCP client. A client connects to the MCP server; the language model is selected inside that client. Only stored, source-backed configurations are shown.

Claude Code

Streamable HTTP

MCP100 logic checked

claude mcp add --transport http exa "https://mcp.exa.ai/mcp"

The publisher connection details were combined with this client's official MCP transport configuration; installation was not executed by MCP100.

Verify safely: Reconnect Claude Code, complete the publisher OAuth or authentication flow if prompted, confirm exa tools are listed, and perform one read-only test.

Uninstall and revoke

Remove exa from Claude Code, clear its stored MCP OAuth credentials, and revoke the publisher authorization if it is no longer needed.

Evidence & methodology

Publisher-documented setup is not the same as MCP100 execution testing. “MCP100 logic checked” means the stored configuration shape was reviewed, not that third-party MCP software was run.

Evidence checked 2026-07-24

Evidence 1 ↗Evidence 2 ↗

Read additional publisher installation details

From the publisher README

Publisher installation details

Original client setup and configuration retained from the publisher documentation.

View original README ↗

Publisher-documented clients

CodexGemini CLIGrokKiroLM StudioOpenCodeWarpWindsurfZed

These names appear in the publisher's installation documentation. MCP100-verified configurations are shown separately above when available.

Installation

Connect to Exa's hosted MCP server:

https://mcp.exa.ai/mcp

Or use a plugin when your client supports one.

Agent Plugin

This repository is an Agent Plugin. Install it with any compatible client.

Claude

Install from the Claude Plugin Marketplace, or run:

claude plugin install exa@claude-plugins-official

Codex / ChatGPT

Install via Plugins in ChatGPT, or run:

codex mcp add exa --url https://mcp.exa.ai/mcp

Other MCP Clients

Most clients can be configured manually with the standard mcpServers shape:

{
  "mcpServers": {
    "exa": {
      "type": "streamable-http",
      "url": "https://mcp.exa.ai/mcp",
    }
  }
}

Exa MCP works with most other clients, point them at https://mcp.exa.ai/mcp.

Client	Where to add it
Kiro	Use the Kiro power, or add manually to `~/.kiro/settings/mcp.json`
LM Studio	Add to LM Studio, or add manually to `mcp.json`
Replit	Add to Replit
Grok Build	`/marketplace` → install Exa, then `/mcp` to sign in
Gemini CLI	Add manually to `~/.gemini/settings.json`
OpenCode	Add manually to `opencode.json`
Windsurf	Add manually to `~/.codeium/windsurf/mcp_config.json`
Google Antigravity	Add manually to `mcp_config.json`
Zed	Add manually to `settings.json` under `context_servers`
Warp	Settings → Agents → MCP servers
v0 by Vercel	Settings → MCP connections

Loading optional publisher installation details…

Show source-catalog configuration

npx -y exa-mcp-server@3.4.0

Capability

What it does

Search the web, retrieve pages, find code, and run research workflows through Exa.

Search & ResearchEducation & ResearchSoftware & Technology

Representative uses

Search the web, retrieve pages, find code, and run research workflows through Exa.
Teams already using Exa that can test with non-sensitive data and least privilege.

Important limits

Users who need an MCP100-recommended recommendation today
Unattended production workflows before permissions and failure behavior are reviewed

Tools2 toolsView tools

web_fetch_exa~156 tokens
Retrieve a web page as clean Markdown.
web_search_exa~250 tokens
Search the web and return ready-to-use results.

Context Footprint

~406 estimated schema tokens

Context impact: Low

Avg. 203 tokens / tool

Estimated token footprint of the MCP tool schemas exposed to the agent.

Runtime verified · 3.2.1 · Aug 21, 2026

Reviewed access

Access & risk

Review the exact access boundary before connecting this MCP to accounts, credentials, or sensitive data.

Authentication: Authentication follows the publisher's documented endpoint or local-server configuration. OAuth scopes, API-key handling, and revocation behavior remain review items.
Permissions: Exact permissions and the read/write boundary have not yet completed MCP100 independent review; consult the linked publisher sources and use the narrowest available scope.
Data handling: Publisher-accessible account or application data may be returned to the MCP client and its selected model. Retention and subprocess behavior require candidate-specific review.
Cost: No MCP100 cost conclusion is published yet. The publisher's account plan, API usage, infrastructure, or per-action charges may apply.
Risk and cautions: Use a test workspace or non-production account, least privilege, read-only controls where available, explicit approval for mutations, and credential revocation after evaluation.

Related reviewed MCPs

Other options

meilisearch

MCP100 Recommended

Inspect and operate Meilisearch indexes and application-search workflows.

Choose this instead if

Choose meilisearch instead if application-search development and diagnosis against Meilisearch.

Inspect and operate Meilisearch indexes and application-search workflows.

Compare with Exa MCP View meilisearch details

opensearch

MCP100 Recommended

Explore OpenSearch indexes, mappings, cluster state, and search results.

Choose this instead if

Choose opensearch instead if search and observability investigation against governed OpenSearch deployments.

Explore OpenSearch indexes, mappings, cluster state, and search results.

Compare with Exa MCP View opensearch details

Elastic MCP

MCP100 Recommended

Connect agents to Elastic search, observability, and security data.

Choose this instead if

Choose Elastic MCP instead if connect agents to Elastic search, observability, and security data.

Expose built-in and custom Elastic Agent Builder tools to external MCP clients.

Compare with Exa MCP View Elastic MCP details

Read publisher documentation

Publisher documentation

Publisher README

Original product, capability, tool, and usage documentation from the MCP publisher. Installation sections appear under Install & Agents.

View original README ↗

Available Tools

Default Tools

Tool	Description
`web_search_exa`	Search the web for any topic and get clean, ready-to-use content
`web_fetch_exa`	Read a webpage's full content as clean markdown from one or more URLs

Optional Tools (enable via the `tools` parameter)

Tool	Description
`agent_run`	Run an Exa Agent for multi-step research, list-building, enrichment, and structured output
`web_search_advanced_exa`	Advanced search with filters, domains, dates, highlights, summaries, and subpage crawling

Enable tools by appending them to the MCP URL (this will replace the defaults, so include all you want):

https://mcp.exa.ai/mcp?tools=web_search_advanced_exa
https://mcp.exa.ai/mcp?tools=web_search_exa,web_fetch_exa,agent_run

Exa Agent requires authentication (OAuth or an API key).

Dynamic Highlights (beta)

Add dynamicHighlights=true to the MCP URL to make web_search_exa and web_search_advanced_exa return Dynamic Highlights. These are highlights chosen across all results under one shared budget, not a fixed budget per URL, so less text reaches your agent's context. When the server runs locally, set DYNAMIC_HIGHLIGHTS=true instead.

https://mcp.exa.ai/mcp?dynamicHighlights=true

Agent Skills

Skills live in skills/ and load with Agent Plugin installs. The Claude plugin in plugins/exa/ ships a copy of them; npm test fails when the copy drifts. Each package version bump on main publishes plugins/exa/ to the root of the claude-plugin branch, which Claude's plugin directory reads. To publish a plugin-only change between releases, run the Publish Claude plugin workflow manually.

Skill	Path	Use when
`search`	`skills/search/`	Deep research, lead gen, competitive analysis, multi-step web investigation
`exa-agent`	`skills/exa-agent/`	Exa Agent runs, enrichment, structured output, Connect providers

Invoke from your client's skill UI (or /skill-name where supported). MCP-only setups still get the tools; skills add orchestration on top.

Authentication

The hosted MCP server works anonymously with rate limits. For higher limits and access to Exa Agent, use either OAuth or an API key.

OAuth is preferred: most clients prompt you to sign in to Exa. To force the login flow (useful for shared connectors and plugins), use https://mcp.exa.ai/mcp?login or https://mcp.exa.ai/mcp/oauth.

If you prefer, you can get an API key from the dashboard and pass it on the URL as ?exaApiKey=…. You can also send it as a Authorization: Bearer … header or an x-api-key header.

Built with ❤️ by Exa

Loading optional publisher documentation…

FAQ

What can Exa MCP search?

Search the web, retrieve pages, find code, and run research workflows through Exa.

Can Exa MCP retrieve full web pages for an AI agent?

The reviewed tool surface includes web_fetch_exa. This does not claim that MCP100 executed those tools.

Does Exa MCP work with Claude Code?

Yes. MCP100 reviewed an evidence-backed setup for Claude Code, but has not independently verified it with a live runtime test.

Is this an official MCP?

Exa is listed as First-party MCP in MCP100’s current evidence.

Current official resources:

Recommended for bounded web and code research when query privacy, result verification, service cost, and downstream use of retrieved content are actively managed.

Best for

Not ideal for

Avoid if

Trade-offs

Evidence

Popularity

Maintenance

Trust review

Sources

Install & Agents

Claude Code

Publisher installation details

Installation

Agent Plugin

Claude

Codex / ChatGPT

Other MCP Clients

What it does

Representative uses

Important limits

Context Footprint

Access & risk

Other options

meilisearch

opensearch

Elastic MCP

Publisher README

Available Tools

Default Tools

Optional Tools (enable via the tools parameter)

Dynamic Highlights (beta)

Agent Skills

Authentication

FAQ

Publisher installation details

Installation

Agent Plugin

Claude

Codex / ChatGPT

Other MCP Clients

Publisher README

Available Tools

Default Tools

Optional Tools (enable via the tools parameter)

Dynamic Highlights (beta)

Agent Skills

Authentication

Optional Tools (enable via the `tools` parameter)

Optional Tools (enable via the `tools` parameter)