Wire it into your agent stack.

• Client

  Claude Desktop

  claude_desktop_config.json

  1. 1Add an entry under "mcpServers".
  2. 2Set the URL to https://api.civilquants.com/mcp.
  3. 3Restart Claude Desktop. The CivilQuants tools appear in the tool picker.
• Client

  Cursor / Windsurf

  MCP settings

  1. 1Open the MCP settings panel.
  2. 2Add a new server: name 'civilquants', URL 'https://api.civilquants.com/mcp'.
  3. 3Save. The CivilQuants tools become available to your agent.
• Client

  Custom agent / SDK

  Streamable HTTP endpoint

  1. 1Connect to https://api.civilquants.com/mcp via the MCP SDK.
  2. 2Free tier requires no auth — rate-limited 10 calls/day per IP.
  3. 3Paid tier: pass your API token in the Authorization header.

• compute_<slug>

  46 assembly compute tools — one per assembly.

  Each assembly is its own tool, e.g. compute_cantilever_wall, compute_manhole, compute_end_area_earthworks. Pass the parameters object and a target standard (free tier: MMHW; paid: CESMM4 / NRM2 / SMM7 / MMHW). Returns BoQ rows, a totals summary, the measured-quantity audit trail, and any geometry advisories. Six assemblies compute on the free tier; the other 40 are paid.

• compute_multi_section_assembly

  One linear assembly, many sections, one combined bill.

  Compute a sectioned/linear assembly (wall, kerb run, drain, barrier) across multiple sections in a single call — each section sets its own run length and geometry, shared parameters apply to all — and get back a combined BoQ plus per-section results. Free tier: MMHW only, free-tier assembly slugs only; each section consumes one daily slot.

• resolve_assembly_from_description

  Plain English in, candidate assemblies out.

  A deterministic router from a job description ("a 3 m deep manhole") to candidate catalogue assemblies — with each assembly's tier, multi-section support, and per-field metadata (unit, min/max, default). It never guesses parameter values and returns no BoQ; the agent fills in the fields, then calls the matching compute tool. Both tiers.

• list_skills · get_skill

  Discover and load the 10 senior-QS skill methodologies.

  list_skills returns every skill's slug, title, summary, and tier (both tiers can browse). get_skill fetches a methodology body — tender review, risk assessment, measurement / contract advice, geotechnical interpretation, earthworks, prelims, pavement, subcontract analysis — which your LLM applies to your own documents on your own tokens. get_skill is paid tier; references are bundled inline.

• get_document_pipeline

  Customer-side chunking + render toolkit for the document-heavy skills.

  Returns the toolkit the document-heavy skills use to chunk a tender pack and render a Word deliverable on your machine — the chunking package, pipeline discipline, and render helpers. Both tiers. Running it over real documents needs a code-execution client (Claude Code, Codex, VS Code); a chat connector can read the toolkit but not execute it.

• render_skill_report

  Render a paid skill's structured output into a branded deliverable.

  After a paid QS skill prompt produces structured JSON, call this tool to render the CivilQuants-branded Word / Excel report and receive a signed download URL. Paid tier.

• save_project · get_project · list_projects · recompute_project

  Persist, reload, browse, and re-measure a multi-assembly project.

  Save a multi-assembly project to your account, fetch it back, list everything saved, or recompute it after changing a parameter — regenerating the per-standard Excel bundle. Paid tier; save_project mints a new project per call (never overwrites). To revise a parameter: get_project, amend, save_project again (e.g. name it R02), recompute — recompute_project takes no parameter overrides. In a new conversation, re-find saved work with list_projects or get_project by exact name.

Connect the server, then paste any of these into your agent. The first three use free assemblies, so they run anonymously with no signup.

• Free tier

  “Using CivilQuants, compute an MMHW Bill of Quantities for a manhole 1.2 m diameter and 3 m deep.”

  An MMHW-structured BoQ plus a dimensioned drawing — every quantity traceable to the formula behind it. (The free tier measures to MMHW; the other three standards are paid.)

• Free tier

  “Size an attenuation tank for a 0.5 ha car park and give me the bill of quantities.”

  The storage BoQ and a drawing for the attenuation assembly.

• Free tier

  “Work out the cut and fill volumes for an embankment cross-section (give the chainage and cross-section areas) using the end-area method.”

  End-area earthworks quantities along the chainage.

• Paid tier

  “Give me the take-off for a 2.5 m high reinforced-concrete cantilever retaining wall, then export it to Excel and a PDF drawing sheet.”

  The wall BoQ plus a downloadable Excel workbook and a PDF drawing sheet.

• Paid tier

  “Save that as a project called “Site A drainage”, then recompute it with the manhole depth changed to 4 m.”

  Your project saved to your account, then re-measured with the new depth.

• Paid tier

  “List the CivilQuants senior-QS skills, then load the subcontract quotation analysis methodology and apply it to these three quotes.”

  list_skills shows the 10 methodologies; get_skill loads the one you pick as system instruction, and your agent runs it against your own documents on your own tokens.

The server ships eleven MCP prompts. One — getting-started — is free and orients the agent. The other ten are senior-QS methodology your LLM loads as system instruction and runs against your own documents, on your own tokens; render_skill_report then turns the result into a branded deliverable. Paid tier.

• getting-started

  Free · guided tour

  Invoke this and the agent asks about your professional context, then proposes the most relevant first call. The only free-tier prompt.

• tender_document_review · tender_risk_assessment

  Paid · tendering

  Senior-QS review of a tender pack and a structured risk assessment, run against your own document on your own tokens.

• qs_measurement_advice · qs_contract_review

  Paid · measurement & contract

  Measurement guidance under the chosen standard and contract-administration review (e.g. variations, payment, claims posture).

• geotechnical_interpretation · geo_environmental_assessment

  Paid · ground

  Design-parameter extraction from a GI report (per BS 5930 + SHW Series 600) and a geo-environmental assessment.

• earthworks_quantities · preliminaries_setup

  Paid · quantities & prelims

  Earthworks quantity methodology and a preliminaries-costing setup the customer's LLM works through on their inputs.

• pavement_design_review · subcontract_quotation_analysis

  Paid · pavement & packages

  Pavement design review and normalised comparison of subcontract quotations (it compares and normalises — it does not price).

Alongside the tools and prompts, the MCP server publishes eight read-only resources at civilquants://*URIs. These are the "skill files" the agent fetches up front to brief itself on the catalogue, the wire schema, and the document-pipeline discipline before producing a report. Cheap to read, idempotent, and edge-cached so they do not count against the daily call budget.

• civilquants://intro

  Agent-onboarding pamphlet

  Short orientation the agent can fetch at the start of a session: who CivilQuants is for, what is in scope vs out, how to think about choosing a standard, free vs paid surface boundaries.

• civilquants://catalogue

  Full catalogue picker tree

  The complete picker tree of every assembly slug and how the catalogue is organised — including categories with no assemblies yet, so the agent knows where extension is coming.

• civilquants://schema/v1

  Wire-model JSON Schema

  JSON Schema for every wire shape — ComputeRequest, ComputeResult, BoQRow, error envelopes, totals summary. Drop into any LLM that supports schema-guided tool use.

• civilquants://document-pipeline-discipline

  Document-pipeline discipline

  Customer-side document-ingestion and chunking pattern for agents that read tender drawings or briefs upstream of CivilQuants. The engine itself ships zero chunking infrastructure at v1.

• civilquants://tools/document-pipeline-starter.py

  Document-pipeline starter script

  A copy-paste Python starter the agent can hand the customer for ingesting and chunking tender documents before feeding parameters into CivilQuants.

• civilquants://tools/document-pipeline-manifest.md

  Document-pipeline manifest

  The manifest describing the document-pipeline pack — what each companion file is for and how they fit together.

• civilquants://tools/chunking-pack

  Chunking pack

  Reference patterns for chunking long tender documents so an upstream agent stays within context while preserving clause-level provenance.

• civilquants://tools/word-report-base.md

  Word-report base template

  The base structure CivilQuants uses for branded Word deliverables — useful context when an agent wants to preview the shape of a rendered skill report.

Tools, resources, and prompts — the three MCP spec primitives — are all live today (see § 04 for the prompts).

• My connector still acts free / anonymous after I subscribed

  The connector is bound to a stale identity from before the subscription. Disconnect and reconnect the CivilQuants connector in your client to re-run OAuth — the fresh token carries your paid tier.

• A tool returned TIER_INSUFFICIENT

  Not an error — the caller is anonymous or unsubscribed and asked for a paid surface (a paid assembly, a non-MMHW standard, get_skill, render_skill_report, or the project tools). Sign up or re-authenticate, then retry the same call.

• drawing or audit_trail came back null

  Withheld, not missing. Trimmed output modes (summary, boq, boq_with_backup) pare the payload for chat contexts; check _meta.output_mode.retrieve_with and re-run with output_mode="full_json" to get the full fields.

• I asked for CESMM4 / NRM2 / SMM7 and got an upsell

  The free tier measures to MMHW only. Naming any of the other three standards requires a paid account — the assembly itself may still be free under MMHW.

• Quantities look wrong on a pipe run / kerb / drain / barrier

  Check the run length first. Linear assemblies scale every quantity directly with it, and an unstated run length (so the default was used) is the most common source of error. Confirm it before computing.

• A deliverable download URL stopped working

  Download URLs are signed and expire after 24 hours. Re-run the compute or render call to mint fresh links — the result is deterministic, so the regenerated deliverable matches.

Still stuck? Email support@civilquants.com with the tool name, the request you sent, and the response envelope you got back.