-
Notifications
You must be signed in to change notification settings - Fork 4
Tools Status
Live functional test results for all 15 MCP tools across normal operation, edge cases, and error handling.
| # | Tool | Status | Auth Required |
|---|---|---|---|
| 1 | web_search |
✅ Operational | None |
| 2 | fetch_page |
✅ Operational | None |
| 3 | reddit_search |
✅ Operational | None |
| 4 | hackernews_search |
✅ Operational | None |
| 5 | arxiv_search |
✅ Operational | None |
| 6 | wikipedia_search |
✅ Operational | None |
| 7 | github_search |
✅ Operational | Optional token or gh CLI |
| 8 | get_github_issue |
✅ Operational |
gh CLI or GITHUB_TOKEN
|
| 9 | groq_search |
✅ Operational | GROQ_API_KEY |
| 10 | groq_analyze |
✅ Operational | GROQ_API_KEY |
| 11 | x_search |
✅ Operational |
AUTH_TOKEN + CT0
|
| 12 | package_info |
✅ Operational | None |
| 13 | package_search |
✅ Operational | None |
| 14 | translate_error |
✅ Operational | None |
| 15 | compare_tech |
✅ Operational | None |
| Edge Case | Tools | Behavior |
|---|---|---|
| Empty query |
web_search, arxiv_search, groq_search, reddit_search, package_search, translate_error, hackernews_search, github_search
|
Returns "Query cannot be empty" (or "Error message must not be empty" for translate_error) |
| Invalid URL |
fetch_page, groq_analyze
|
Error with UnsupportedProtocol or graceful message |
| Nonexistent page/issue |
wikipedia_search, arxiv_search (Lucene), get_github_issue, package_info
|
Clean "not found" error |
| Unparseable input | translate_error |
Returns "Could not identify the error type or programming language" with guidance |
| Special characters | web_search("!@#$%^&*()") |
Works normally |
| Single character query | web_search("a") |
Works normally |
| Very large page | fetch_page |
Truncates gracefully with warning |
| Invalid domain scope | web_search(domain="...") |
"No results found" |
Package names without dots or @ default to npm registry. For PyPI packages, specify explicitly:
package_info(name="requests", registry="pypi")
// Without registry="pypi", defaults to npm (v0.3.0, 11.9K/wk)
// With registry="pypi", returns v2.32.3, 212.4M/wkThe tool uses an auto backend that tries httpx first, then falls back to curl with TLS impersonation if blocked. Some sites (Cloudflare, aggressive bot detection) may require backend="curl" explicitly.
The arXiv export API (export.arxiv.org) occasionally returns HTTP 503 during maintenance. This is an upstream issue, not a code defect. Retry after a few minutes.
Works best with well-known error patterns (AttributeError, TypeError, CORS errors, etc.). Generic or nonsensical input returns a "could not identify" message rather than searching Stack Overflow.
Home | Tools | Configuration | Architecture | Development