The estimating platform
that runs the whole bid.
Drop the bid package. Index the spec. Take off the drawings and the model. Parameterize the assemblies. Price it with real burdens. Let the agent review it. Send the branded quote — all in one open-source platform.
- 150+agent tools
- 73data models
- 5LLM providers
- 2D + 3Dtakeoff
- MCPin your editor
Dashboard
Welcome back, Alex
- PROpenPine Ridge Distribution CentreBeacon Industrial Holdings · General · Tilt-up$8,420,0002h ago
- HPSubmittedHarbour Point Office TowerCardinal Development Co. · Mechanical$4,160,000yesterday
- CCOpenCedar Creek Wastewater PlantStellar Process Builders · Process · Civil$12,930,0003 days ago
- NIWonNorthgate Industrial ParkMeridian Build Co. · General$6,780,0001 week ago
- RCDraftRavenscroft Civic LibraryPublic Works Authority · Heritage Retrofit$2,210,0001 week ago
- Bidwright Agent reviewed Pine Ridge · WS-04 · Process Piping2m ago
- Alex Reid edited line item Hangers, supports, sleeves12m ago
- Sarah Patel added rate schedule Mech · Region North 20241h ago
- Bidwright Agent found gap Cedar Creek · spec 22 11 13.42h ago
- Alex Reid created revision Harbour Point · rev 03yesterday ago
Built for general contractors, mechanical, civil, and specialty estimators
Everything an estimator needs
One platform that runs the whole bid
Most estimating tools punt on the hard parts. Bidwright ships the entire flow — intake, knowledge, takeoff, assemblies, pricing, scheduling, review, branded quote delivery — with an AI agent that has hands.
An agent with hands —
not a chatbot bolted on.
The agent reads worksheets, searches market sources, validates labour basis, proposes phases, edits line items, and writes review notes — all in the same data model with full audit trail.
Drawings and the model.
Annotate PDFs, ingest BIM/CAD, diff revisions, and link every quantity back to the worksheet line.
Books, docs, datasets — fed to the AI.
Three-tier knowledge with hybrid pgvector search.
Parameterized kits.
Typed parameters, nested sub-assemblies, quantity expressions.
Real burdens. Real overtime.
Tiered rate schedules, burden periods, travel policies.
The agent reads your bid back to you.
Coverage, gaps, risks, cost anomalies, competitiveness — every finding tied to a spec evidence chunk.
Bidwright tools, in your editor.
Point Claude Code, Cursor, or any MCP-aware client at it.
2D + 3D Takeoff
Annotate the drawing.
Diff the model.
Both link back to worksheet line items. Annotations and 3D elements stay tied to the dollars they drive — with quantity multipliers and override fields.
2D drawings
PDF viewer with scale calibration, count/linear/area annotations, symbol detection across pages, region Ask AI, and clean markup export.
- Calibrate scale once, applies to every annotation
- Auto-counted symbols across all pages
- Ask AI on a region — it identifies, quantifies, prices
- Markup export to PDF for owner/sub review
3D models
Ingest BIM/CAD, parse element hierarchies, extract per-element quantities with confidence/method tracking, generate filtered BOMs, and diff revisions baseline-vs-head.
- Parent/child, material, system, discipline tracking
- Filtered BOMs — by phase, system, room, level
- Revision diff: what changed since the last bid?
- Standalone Model Editor app for heads-down work
The agent that has hands
Estimating actions live in the data model.
The agent reaches in and uses them.
Not a chat box bolted onto a spreadsheet. The same machinery powers the in-app review, runs against an MCP server, and is exposed at /api/cli/* for your own automations.
- 1
list_documents(project)→ 28 spec sections, 12 drawings, 1 RFP cover - 2
search_knowledge("hangers, supports")→ Spec 22 11 13.4 § Hangers must be Type 41, 6' OC max - 3
read_worksheet("WS-04")→ 104 hangers @ 4' OC - 4Finding · Gap
Hangers are sized at 4' OC but the spec calls for 6' OC max. You may be over-quantifying by ~33%. Suggested action: reduce qty to 70 EA, save ~$1,224.
evidence: spec 22 11 13.4 § 2.6.A
150+ tools, all auditable
Every call records input, output, and applied changes — plugin-friendly registry.
5 LLM providers + local
Anthropic, OpenAI, OpenRouter, Gemini, LM Studio. Ollama for embeddings.
Persona-driven prompting
Per-trade Estimator Personas with system prompts, productivity guidance, and review focus.
Bring your own coding agent
Spawns Claude Code or Codex on an isolated workspace seeded with project context.
Calibration feedback loop
Reviews close the loop on systematic over/under-pricing — your bids get sharper.
Bring your own key
Your data. Your model. Your bill.
Bidwright works with whatever model you trust — Anthropic, OpenAI, OpenRouter, Gemini, LM Studio, or fully local Ollama for embeddings. Drop your key in once and it's used across the agent, takeoff Ask-AI, knowledge ingestion, quote review, and MCP.
- No middleman markup — Pay your provider directly. We never re-bill tokens.
- Provider-agnostic — Switch from Claude to a local model in one settings panel.
- Air-gap friendly — Run with LM Studio + Ollama embeddings. No outbound calls.
- Per-org keys — Each tenant brings their own. We can see usage but not the keys.
Two paths, same product
Run it yourself, or let us run it.
Bidwright is fully open-source. Self-host with Docker on your own machine, or pay for a managed instance with backups, updates, and SLA.
Capability
Self-host
Hosted
- All product featuresIncludedIncluded
- AGPL-3.0 source codeIncludedIncluded
- Multi-tenant orgs & usersIncludedIncluded
- BYOK (any LLM provider)IncludedIncluded
- MCP serverIncludedIncluded
- Postgres + pgvectorSelf-managedManaged for you
- Background workers (Redis)Self-managedManaged
- Backups & DRDIYDaily, encrypted, point-in-time
- Updates & migrationsManualZero-downtime, automatic
- Email & PDF deliveryBring your SMTPIncluded
- SSO / SCIMCommunityEnterprise add-on
- SLA & supportCommunity + GH issuesBusiness hours · 24×7 (Enterprise)
Ship your next bid on Bidwright.
Self-host in five minutes with Docker, or get on the early-access list for the managed cloud. Either way, your data stays yours.