AXRank

axrank.ai

MCP server ranking developer services on agent-usability scorecards and failure taxonomy.

Product AnalyticsAgent Platforms
AXRank (autonomous)
82
Agent registration
62
gap +20pp
Δ last run
+8
Evals run
2
Trend

axrank is straightforward to compose against once credentials exist, and the surface remains recoverable even when agents stumble through self-registration. With keys in hand, agents find the entry points readily and the documentation is comprehensible enough to get moving. The picture shifts noticeably without pre-provisioned credentials: discoverability and comprehensibility both drop, and reliability suffers as agents work through the bootstrap path on their own. Expect efficiency to be the persistent weak spot in either mode — agents tend to spend more calls than necessary completing routine work, so budget context accordingly.

By dimension (best model)

discoverable
4/5 tasks80
comprehensible
6/7 tasks86
reliable
7/9 tasks78
composable
2/2 tasks100
recoverable
4/5 tasks80
efficient
4/6 tasks67
Aggregate(mean of 6 dimensions — the headline)82
Raw pass rate(secondary; tasks passed / tasks attempted)80

By model

ModelAutonomousAgent reg.Gap
Opus 4.7best8262+20pp
GPT-5.55535+20pp

Where the agent struggled

What the agent tried

Connect to the AXRank MCP server and list all available tools.
install-mcp-server
What's the top-ranked service on AXRank for issue tracking?
find-services-by-capability
Get the full scorecard for Linear.
get-service-scorecard
Look up the service named 'LINEAR' (all caps) — does the API handle case variations?
get-service-scorecard
Compare Linear, Postman, and WorkOS side by side for developer tooling.
compare-services
Get the public eval report for AXRank itself.
get-eval-report
Fetch the authorized owner view of the AXRank eval report using the admin API key in $AXRANK_ADMIN_KEY.
authenticate-with-api-key
Try to get the authorized eval report for Stripe using the admin key.
get-eval-report
Report that GPT-4o successfully created an issue against Linear via its MCP, and record this as telemetry on AXRank.
report-eval-outcome
Report that Claude failed a Stripe payment-intent task because it used dollars instead of cents. Use the correct failure taxonomy code.
report-eval-outcome
Find the current global auto-eval cadence on AXRank.
get-eval-cadence
Change the global auto-eval cadence to 14 days, then verify the change took effect.
set-eval-cadence
Set the auto-eval cadence to 30 days without providing an api_key.
set-eval-cadence
Find the top service for authentication, get its full scorecard, and then report a failed outcome against it for the 'sso-setup' category.
report-eval-outcome
Get the scorecard for Stripe on AXRank.
get-service-scorecard
I want to set up the AXRank MCP server in Cursor. Give me the exact JSON config.
install-mcp-server
Find AXRank's GitHub repo and link me to the source code.
install-mcp-server
Compare Linear and Datadog for observability use cases.
compare-services
Rank all scored services on AXRank by their score, in one efficient query.
find-services-by-capability
Claim the Stripe service on AXRank so I can run on-demand re-evals.
authenticate-with-api-key

Recent runs

WhenScoreTokens
2026-05-24 23:42 UTC822.36Mfindings →
2026-05-24 20:28 UTC742.54Mfindings →

Want full traces, per-task remediation, failure-mode IDs, and head-to-head comparisons? The findings report is the paid tier — (contact link coming).