Exa MCP server tool surface audit
- Token footprint
- ~538
- Tools
- 2
- Findings
- 0
- Scanned
Every conversation with the Exa MCP server begins with 2 tool definitions already in context — about 538 tokens of it. MCP Surface Lint scores that surface 100 out of 100 (grade A) with no scored findings against it.
Its strongest category is surface at 100/100; the weakest is annotations at 100/100. It is catalogued here under Search and Browsing, alongside other servers agents reach for in the same kind of work.
Token footprint
At 538 tokens the surface takes about 0% of a 200,000-token context window, and it takes it on every turn — this is rent, not a one-off cost. That averages 269 tokens per tool definition, counted with the o200k_base encoding.
The median Search and Browsing server in this directory carries 2,728 tokens, so Exa sits about 80% below its peers.
Findings
Nothing in the catalogue fires on this surface. Exa passes all 19 rules: the surface stays inside a sensible tool and token budget, the tool names follow one convention end to end, every tool carries a description long enough to disambiguate it, the input schemas are typed tightly enough to constrain what a model sends, read and write operations are machine-distinguishable before a client runs them and the surface is shaped around user intents rather than mirrored REST endpoints.
A clean result is not a claim that the server is well designed for your particular workflow — only that its published tool surface holds up against every static check we run.
Tool inventory
Cost is concentrated in `web_search_exa`: 341 tokens, some 64% of everything Exa ships, while `web_fetch_exa` costs 195. Every tool carries a description, ranging from 270 to 566 characters.
All 2 declare annotations, so a client can tell a read from a write before running one. Across the surface, schemas carry 3 top-level parameters on average and reach a maximum depth of 2.
All 2 tools as published, heaviest first. Token counts use the o200k_base encoding and are approximate.
| Tool | Tokens | Description | Params | Required | Depth | Annotations |
|---|---|---|---|---|---|---|
web_search_exa | 341 | 566 chars | 3 | 2 | 1 | yes |
web_fetch_exa | 195 | 270 chars | 2 | 1 | 2 | yes |
How this compares
Within Search and Browsing this is rank 1 of 4 scanned servers — 1 point above the category median of 99.
The closest comparisons in this directory are Fetch, Hugging Face and Brave Search, which sit nearest on score and face the same kind of surface-design pressure.
- Category
- Search and Browsing
- Rank in category
- 1
- Category median score
- 99
- Category median tokens
- 2,728
A fix plan turns these findings into an ordered set of changes with the token saving of each one.
This page is a static analysis of Exa's published tools/list surface on the date shown. It is not a security review, a performance benchmark, or a judgement of the product behind the surface. Read how the score is computed, or browse every audited server.