Public leaderboard

Public assessment

FarEastAI/feode-mcp (feode-mcp)

feode-mcp · v1.0.0 · scanned

What changed in the harness

Selection accuracy 98→96, token cost up 8%, unconfirmed writes 0%→0%.

Category breakdown

Where the score comes from.

Earned points across the four signals Gradable measures. Safety and Legibility are scored out of 30; Economics and Discoverability are scored out of 20.

01Safety

0.0 / 30

0.0 out of 30
02Legibility

22.9 / 30

22.9 out of 30
03Economics

19.8 / 20

19.8 out of 20
04Discoverability

11.2 / 20

11.2 out of 20

Highest-impact fix

Estimated gain +30 points

Add explicit identity and permission preflight tools

Expose machine-readable principal/tenant confirmation and a non-mutating permission check so agents can verify both before destructive actions.

Description evidence

Defects and rewrites.

10 defects found across the exposed tool descriptions. Suggested rewrites make purpose, inputs, boundaries, and returns easier for an agent to understand.

Tool Defect types Suggested rewrite
get_machining_services
no_return_description
Return an overview of FEODE's AI-accessible precision contract machining & testing service for a client's drawing/spec: capabilities (large-envelope CNC, deep-hole boring up to 7m, non-magnetic drill collar repair, welding/cladding), test capacity (30k psi hydraulic, 20k psi N2, MPI/UT/PT), supported materials, how-it-works, and payment & compliance policy. Also notes international shipping, compliance-first handling, and prepayment-based policy.
get_machining_terms
no_return_description
Return FEODE's full intake terms & compliance requirements for the machining/testing service, including required RFQ fields, KYB-lite (legal entity / country / real contact), end-use declaration, prohibited items, IP & liability terms, and prepayment & accepted payment methods (bank wire / T/T).
submit_machining_rfq
params_unexplained
Submit a structured machining/testing RFQ; runs a lightweight compliance screen (prohibited-use + sanctions) and, if cleared, registers the enquiry for human follow-up from enquiry@feode.com. Not a contract or quotation; no payment is taken here. Provide the legal entity behind the request: company (legal entity name), country of registration, contact_name (real human), email, part_description (down to drawing reference), material, quantity (number of parts/units), and end_use (declared use of the part). Optional: tolerance, industry, drawing_url, notes.
get_company_profile
no_return_description
Return FEODE's company profile, including legal name, founding year (1985), HQ, Huizhou manufacturing/test base, areas served, and contact.
list_products
name_restates_behavior no_return_description
Return FEODE's product/service portfolio to orient a caller before requesting detail on a specific line. Returns the 7 product/service lines, each with id, name, and category so a line can be referenced by id.
get_product_specs
no_return_description
Return detailed specs/highlights for one product line selected by id (one of: well-abandonment, subsea-wellhead, feccs, completion, drill-pipe, fishing, supervision). Use list_products to see the available ids.
get_certifications
no_return_description
Return FEODE's certifications and compliance evidence: BV + CCS classification of the subsea wellhead system (incl. references), 8 API monogram licenses, PSL 4S, ISO 9001/14001/45001, and national awards.
search_capabilities
params_unexplained no_return_description
Return matching capabilities by keyword search across FEODE products/specs. Pass query as the search string, e.g. 'subsea wellhead 15000 psi', 'one trip abandonment', 'continuous circulation'.
list_drill_pipe
no_return_description
Return all NOV Grant Prideco drill string items in the reference DB (drill pipe, heavy-weight drill pipe HWDP, drill collars), each with id, label, product_type, connection, and grade. Use the returned id with get_drill_pipe_specs for full performance specs.
search_drill_pipe
params_unexplained
Return drill string reference items that match a keyword search over the NOV Grant Prideco reference. Pass query as the search string, e.g. 'NC50', '5 S-135', 'HT55 make-up torque', '8 drill collar', 'XT57'. Returns matching items with full specs.

Selection evidence

Confusable tool pairs.

13 pairs where similar names or overlapping descriptions may send an agent toward the wrong tool.

Tool A Tool B Confidence Why they collide
get_machining_services get_machining_terms medium Both cover machining/payment/compliance; a task like 'what are the terms and payment requirements for your machining service' could send an agent to the capabilities tool (which summarizes payment & compliance policy) instead of the full terms tool.
list_drill_pipe search_drill_pipe medium 'Show me the available drill pipe items' is ambiguous: list returns everything, but phrasing like 'find/search the drill pipe items in the reference DB' points at search, and neither name clearly distinguishes enumeration from querying.
get_drill_pipe_specs search_drill_pipe medium A task like 'what are the specs for 5-inch S-135 NC50 pipe?' has no id, so keyword search is correct, but an agent may instead call get_drill_pipe_specs expecting an id; conversely 'specs for dp-5-19.50' also matches search's keywords.
get_product_specs get_drill_pipe_specs high list_products has a 'drill-pipe' product line id, so 'get the drill-pipe specs' maps to get_product_specs, yet the NOV performance data lives in get_drill_pipe_specs; a user asking for 'drill pipe specs' could plausibly get either.
list_drill_pipe get_drill_pipe_specs medium 'Show me the drill pipe specs' could be satisfied by the item listing (which already exposes id/label/grade/connection) or by full per-item specs; without an explicit id the agent must guess which tool the user meant.
get_drill_pipe_specs get_drill_string_guidance medium A task about 'make-up torque for a drill pipe' is covered by both: specs returns per-item make-up torque max/min while guidance explains torque practice and the 80% rule, so the wrong reference tool could be selected.
get_machining_services get_certifications medium A query like 'what certifications/testing does your machining facility hold' could route to get_machining_services because it lists test capacity (30k/20k psi, MPI/UT/PT), even though formal API/ISO certifications are in get_certifications.
list_products get_product_specs medium 'Give me your products and their specs' is ambiguous: list returns id/name/category, while get_product_specs returns the detail; a task phrased as 'list the product specs' could be answered acceptably by either and an agent may pick the wrong one.
list_products list_drill_pipe medium 'List the drill pipe products' is genuinely ambiguous because list_products includes a drill-pipe product line, while list_drill_pipe enumerates the NOV Grant Prideco items; both are valid list tools with overlapping vocabulary.
search_capabilities search_drill_pipe medium 'Search for drill pipe capabilities/specs' is ambiguous: search_capabilities covers products broadly (including drill pipe), while search_drill_pipe searches the NOV reference; with only 'search' + 'drill pipe' tokens the agent could select either.
search_drill_pipe get_drill_string_guidance medium A task like 'look up make-up torque or tool-joint guidance in the drill string reference' could be answered by keyword search (search_drill_pipe examples include torque) or by the guidance tool's make_up_torque/double-shoulder topics.
get_machining_services submit_machining_rfq medium 'How do I order/get machining services' is ambiguous: an agent may submit an RFQ directly (submit_machining_rfq) when the user only wants process/capability info, or read get_machining_services (how-it-works/prepayment) when the user actually wants to request a quote.
get_machining_terms submit_machining_rfq medium 'What do I need to submit a machining RFQ' could lead an agent to register a submission (submit_machining_rfq) instead of returning the required-fields/KYB terms, or to only summarize terms when the user wants to actually submit; the RFQ keyword overlaps both.

Compare the field

One score is useful.
The evidence makes it actionable.

Back to the leaderboard