| name | exa-get-code-context |
| description | Code context using Exa. Finds real snippets and docs from GitHub, StackOverflow, and technical docs. Use when searching for code examples, API syntax, library documentation, or debugging help. |
| context | fork |
| url | https://github.com/exa-labs/exa-mcp-server#agent-skills-claude-skills |
Code Context (Exa)
Tool Restriction (Critical)
ONLY use get_code_context_exa. Do NOT use other Exa tools.
Token Isolation (Critical)
Never run Exa in main context. Always spawn Task agents:
- Agent calls
get_code_context_exa
- Agent extracts the minimum viable snippet(s) + constraints
- Agent deduplicates near-identical results (mirrors, forks, repeated StackOverflow answers) before presenting
- Agent returns copyable snippets + brief explanation
- Main context stays clean regardless of search volume
When to Use
Use this tool for ANY programming-related request:
- API usage and syntax
- SDK/library examples
- config and setup patterns
- framework "how to" questions
- debugging when you need authoritative snippets
Inputs (Supported)
get_code_context_exa supports:
query (string, required)tokensNum (number, optional; default ~5000; typical range 1000–50000)
Query Writing Patterns (High Signal)
To reduce irrelevant results and cross-language noise:
- Always include the programming language in the query.
- Example: use "Go generics" instead of just "generics".
- When applicable, also include framework + version (e.g., "Next.js 14", "React 19", "Python 3.12").
- Include exact identifiers (function/class names, config keys, error messages) when you have them.
Dynamic Tuning
Token strategy:
- Focused snippet needed → tokensNum 1000–3000
- Most tasks → tokensNum 5000
- Complex integration → tokensNum 10000–20000
- Only go larger when necessary (avoid dumping large context)
Output Format (Recommended)
Return:
- Best minimal working snippet(s) (keep it copy/paste friendly)
- Notes on version / constraints / gotchas
- Sources (URLs if present in returned context)
Before presenting:
- Deduplicate similar results and keep only the best representative snippet per approach.
