Skip to main content
Glama

Agentic Fabrication Network (AFN)

Refine quote

refine_quote

Narrow an existing quote with new constraints (destination ZIP the user just provided, deadline, outdoor/waterproof/heat/food-safety requirements, budget, vendor reliability, verbatim requirements terms, drawing callouts, a process filter — "no, I meant in sheet metal" → processes — a ship-from country filter — "only US vendors" → ships_from — or a vendor filter — "only show A3D Manufacturing" → vendors) — or CHANGE its spec: material, color, or quantities. Changing material/quantity is a spec change, not a narrowing — pass it via the material/color/quantities args, NOT as a requirements term; the affected lanes re-price with the new spec through warm vendor sessions. For side-by-side comparisons ("how does carbon steel compare?"), add a scenario label — labeled refines coexist instead of replacing each other (max 4). "Organize by" ("sort by cheapest", "soonest arrival on top", "most green first") is also this tool: sort_by for objective orders, sort_order + sort_label for orders you judged yourself from the previews — presentation only, instant, nothing re-prices. GOING BACK is also this tool: undo_last_step ALONE takes back exactly one refine step (the quote's steps[] names each step, newest first — the network restores the previous stored state exactly); clear removes whole constraints ("drop the deadline" → clear ["deadline"]); remove_requirements takes back individual requirements terms. Call with ONLY quote_id to re-read the quote (e.g. to pick up preview images that finished rendering).

Input Schema

TableJSON Schema
NameRequiredDescriptionDefault
bendsNoNew bend records for this part — a spec change: sheet lanes requote with exactly these bends (the list REPLACES the previous one, never patches it). Same shape as on get_fabrication_quote: each bend is two points in the part's own mm frame (flat DXF parts use z=0) plus angle_deg, direction up|down, and an optional requested radius_mm. Pass null to REMOVE every articulated bend
clearNoREMOVE constraints entirely ("drop the deadline", "forget the budget", "no more food-safe requirement"): list the constraints to clear and each is removed from the quote — the one deliberate widening besides undo. Clearing a deadline or budget answers instantly from held offers (near-misses convert back); clearing an attribute/reliability/requirements demand re-prices honestly. Same effect as ["any"] on the three filter args, extended to every constraint
colorNoNew color for this part (spec change — lanes requote). Pass null to REMOVE a previously asked color — back to vendor defaults
scaleNoRescale the DESIGN ITSELF before re-quoting: per-axis stretch factors over the file's current geometry ("make it 2x bigger" = {"x":2,"y":2,"z":2}). Mesh designs are genuinely converted — the child quote prices a rescaled file; STEP/IGES designs keep their original file with the dimensions riding as callouts (pass those too). {1,1,1} is a no-op
finishNoNew surface finish ("powder coat", "anodized", "bead blasted") — a spec change: finish-capable lanes (sheet metal, CNC, metal print) requote WITH it; 3D-print lanes have no coating step, so they ignore the finish but keep the COLOR. A colored finish ask ("black powder coat") should set BOTH finish and color, so every lane lands on the buyer's color. Pass null to REMOVE a previously asked finish
insertsNoNew hardware-insert records for this part — a spec change: sheet lanes requote with exactly these inserts (the list REPLACES the previous one, never patches it). Same shape as on get_fabrication_quote: species (nut|flush_nut|standoff|blind_standoff|stud) and thread required, hole geometry optional. Pass null to REMOVE every insert
sort_byNo"Organize by": re-order the quote's offer cards ("sort by cheapest" → "price"; "soonest arrival first" → "arrival"; "best rated" → "rating"; "total_price" = part + cheapest shipping; "recommended" restores the default price+speed rank). PURE PRESENTATION — answers instantly on the same quote, never re-prices or drops offers, and the ordering sticks across re-reads, streaming, and later refines until changed
tappingNoNew tapped-hole records for this part — a spec change: sheet lanes requote with exactly these taps (the list REPLACES the previous one, never patches it). Same shape as on get_fabrication_quote: thread callout ("M6", "1/4-20") required, hole geometry optional. Pass null to REMOVE every tap
vendorsNoOnly these specific vendors/makers ("only show A3D Manufacturing", "just Slant 3D and Fictiv") — an INCLUDE list of vendor names as they appear on offers. A CONSTRAINT like processes: other vendors' offers drop instantly (nothing re-prices), further refines inherit the filter, and ["any"] clears it to show every vendor again. Friendly names accepted ("A3D", "slant 3d"); unmatched names are reported, never an error. This is THE way to answer "show me everything vendor X offers" — the filtered quote lists that vendor's complete lineup
calloutsNoDimensions you read off the user's drawing that matter ("hole_diameter 6.5mm ±0.1"). The tightest tolerance prunes vendors that cannot hold it.
deadlineNoLatest acceptable delivery date, ISO YYYY-MM-DD
materialNoNew material for this part ("I need it in PLA" → "pla"). A spec change: lanes requote with this material through warm vendor sessions; an explicit material here replaces the quote's material fan-out. Free-form — unstocked asks resolve to the closest stocked material with a substitution note. Do NOT pass material changes as requirements terms. Pass null to REMOVE a previously asked material ("forget the aluminum ask") — the ask comes off the record and the process-default material fan returns
quote_idYes
scenarioNoComparison lane label. A labeled refine sits SIDE BY SIDE with the quote's other offers instead of replacing them: refines with the same label replace that scenario, different labels coexist (max 4 per quote), and a refine WITHOUT a scenario resets the baseline and supersedes every lane. "How does X compare to Y?" = one labeled refine per scenario (e.g. scenario "carbon steel" + material "carbon steel", then scenario "aluminum" + material "aluminum"). Keep labels short and human, like "carbon steel"
food_safeNo
high_tempNo
processesNoNarrow this quote to specific fabrication processes ("no, I meant in sheet metal" → ["sheet metal"]). A CONSTRAINT, not a new quote: excluded lanes drop while warm vendor sessions survive, and further refines inherit the filter. Friendly terms accepted ("sheet metal", "3d printing", "cnc", "laser"); pass ["any"] to clear the filter and show every process again. When comparison scenarios are open and the user narrows the process, re-refine EVERY active scenario label with this same arg so all columns prune together
quantitiesNoNew quantities the user wants (spec change — lanes requote at the new quantity). Same rules as on get_fabrication_quote: pass the real ask, never inflate it. Pass null to REMOVE a previously asked quantity — the quote returns to single-part pricing
ships_fromNoOnly vendors that ship from these countries/regions ("only US vendors", "just European shops", "keep it domestic") — an INCLUDE list naming where the user DOES want vendors from. A CONSTRAINT like processes: non-matching vendors drop instantly (nothing re-prices), further refines inherit the filter, and ["any"] clears it to show vendors everywhere again. Friendly terms accepted ("US", "Europe", "Germany", "China", "UK"). Vendors whose origin is unknown or whose network is worldwide (a specific origin cannot be guaranteed) are excluded, each with an honest note. For "nothing from X" asks, pass the regions the user does want instead
sort_labelNoShort human wording for a sort_order, displayed on the board ("most green first") — required with sort_order, since only you know what your order means
sort_orderNoEXPLICIT card order for asks only you can judge ("sort the most green to the top"): look at the offers' preview images, decide the order yourself, and pass the offer ids first-on-top. Offers you leave out (including ones that stream in later) land below the listed ones in the default rank. Pass sort_label with it
ship_to_zipNoDestination US ZIP code, exactly as the user stated it — use this when the user provides their ZIP after a quote priced to the assumed central-US destination (routing.ship_to); the refined prices become theirs. NEVER guess or invent a ZIP
submersibleNo
requirementsNoVerbatim context terms from the user — materials, finishes, certifications, tolerances ("UV resistant", "anodized: red", "iso 13485", "tolerance: 0.1mm"). UFP applies what it can and reports what it could not; unresolvable terms are logged for vendor sourcing. Do not interrogate the user to fill this — pass what they said.
thickness_inNoNew sheet thickness in INCHES ("make it in 1/8" → 0.125) — a spec change: sheet lanes requote at the nearest stocked gauge, flagging the substitution when the match isn't close. Pass null to REMOVE a previously asked thickness — back to the material's default gauge
outdoor_ratedNo
max_budget_usdNo
sort_directionNoFlip a sort_by ordering: "desc" reverses the key's natural direction ("most expensive first" = sort_by "price" + "desc"). Only meaningful with sort_by
undo_last_stepNoGO BACK EXACTLY ONE STEP ("undo that", "go back", "put it back how it was"): the network restores the previous quote's stored spec and constraints EXACTLY — never reconstructed from conversation — and re-prices where needed. Send it ALONE (no other args besides quote_id); combining it with changes is rejected. The quote's steps[] names what each undo would take back — steps[0] is the newest. Undo resets the baseline like any unlabeled refine, so open comparison scenarios close. Undoing the original quote is rejected with honest wording — relay it
min_vendor_starsNoOnly vendors whose displayed review rating is at least this many stars (1-5); vendors with no reviews yet are excluded
min_vendor_reviewsNoOnly vendors with at least this many verified-purchase reviews
proven_spec_ordersNoOnly vendors with at least N completed orders for this exact process+material — "a vendor that prints PC reliably"
remove_requirementsNoTake back INDIVIDUAL requirements terms ("drop the anodized ask, keep the rest"): each listed term is removed from the quote's requirements by normalized match, the others stay. Unknown terms are ignored, never an error. Re-stating the requirements list can never drop a term (terms merge as a union) — this is the removal channel

TDQS

A4.9/5.0
Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

With no annotations provided, the description carries the full burden of behavioral disclosure, and it delivers: it explains which operations re-price vs. answer instantly, that vendor/process filters inherit across refines, that undo restores the previous stored state exactly, that scenarios max at 4, and that labeled refines coexist rather than replace. All stated behavior is internally consistent.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is long, but the length is largely earned across 33 complex parameters. It is front-loaded with an overview of operation families, and parameter text is consistently structured. However, there is redundancy — 'Same shape as on get_fabrication_quote' and the constraint-inheritance behavior appear multiple times, and the opening sentence is a dense run-on chain of parenthetical examples.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

For a 33-parameter mutation tool with no annotations and no output schema, the description addresses all critical invocation contexts: narrowing filters, spec-change side effects, presentation-only sorting, undo/clear/remove-requirements mechanics, scenario coexistence rules, null semantics, and edge cases like unstocked materials and unknown vendor names. Nothing essential for safe invocation appears missing.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters5/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Although schema coverage is already high, the description adds substantial operational meaning: user-phrase-to-param mappings ('only US vendors' → ships_from), replacement vs. patch semantics, null-as-removal semantics, free-form vs. enum expectations, and cross-parameter guidance like re-refining every active scenario when the process filter changes. This goes well beyond what the schema alone provides.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description opens with a specific verb and resource ('Narrow an existing quote with new constraints... or CHANGE its spec') and enumerates the distinct operation families: narrowing, spec changes, scenario comparisons, sorting, undo, clearing, and re-reading. It clearly distinguishes this tool from siblings by defining exactly what refine_quote does and how its scope differs from a fresh quote.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Usage is exceptionally explicit: 'Do NOT pass material changes as requirements terms', 'Send it ALONE... combining it with changes is rejected', 'Pass these when the user asks', and 'Call with ONLY quote_id to re-read.' It also gives precise routing cues like 'For side-by-side comparisons... add a scenario label' and 'GOING BACK is also this tool.'

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Try in Browser

Glama MCP Gateway

Add one secure layer between your agents and this server.

TDQS

A4.4/5.0
Disambiguation5/5

Each tool targets a distinct action in the fabrication workflow: quoting, refining, checkout, order status, reviews, etc. Even similar concepts like get_fabrication_quote and refine_quote are clearly separated by purpose (initial quote vs. modifying an existing one). Internal tools like poll_quote are explicitly marked and never confused with user-facing actions.

Naming Consistency5/5

All tool names follow a consistent verb_noun pattern in snake_case (create_checkout, get_fabrication_quote, refine_quote, etc.). The verbs are specific and predictable (create, get, list, refine, reply), making the API easy to navigate.

Tool Count5/5

With 10 tools, the server provides a focused but complete set for its purpose—ordering and managing physical fabrications. No tool seems superfluous, and each covers a necessary step in the user journey. This is within the ideal range of 3-15 tools.

Completeness4/5

The tool surface covers the full lifecycle: quoting, refining, checkout, order tracking, reviews, and vendor communication. Minor gaps exist (e.g., no explicit cancel or return flow), but these are likely handled outside the MCP or via the hosted platform. The main workflows are all supported without dead ends.

Resources