AI agents

Run BuildWorkPro from the agent you already talk to

Connect Claude, ChatGPT, Grok Bot, OpenClaw or Hermes Agent once. Then describe the job, forward the email, ask the question. The bid, the lead, the change order, the pay app land in your account with your permissions, and you review before anything goes out.

Office Bot Grok Bot · always on
  1. Morning. Two GC emails overnight look like real work. Created leads for the storefront job and the curtain wall RFI, contacts linked. 6:40 AM
  2. The storefront GC wants a walkthrough Thursday 9 AM. It’s on your calendar. Bid due the 19th, set on the lead. 6:41 AM
  3. Draft the walkthrough confirmation and ask about the glass spec. 7:05 AM
  4. Draft is in your inbox for review. Logged it on the lead. 7:05 AM
Message Office Bot
app.buildworkpro.com/leads
BuildWorkPro sales pipeline with new leads created by an AI agent

"When I start a new bid, I usually just start chatting with Claude on my computer or my phone. I give it the information and it builds the bid for me. When it's done, I review it in BuildWorkPro and send it. Projects work the same way — the team and I run almost everything through natural conversation now."

— Ivan, founder of BuildWorkPro and owner of National Glass and Construction

How it works

  1. 1

    Connect once

    Paste one address into your agent and sign in to BuildWorkPro. A consent screen shows exactly what the agent is asking for before you approve. Install guides for every agent are below.

  2. 2

    Talk like you'd talk to your estimator

    "New bid for the storefront job — twelve openings, our standard hardware, usual labor rates." Or forward the GC's email and say "make this a lead." The agent creates the record, adds the line items, prices them from your catalog.

  3. 3

    Review and send in BuildWorkPro

    Everything the agent builds lands in your account looking like you made it, because it did it with your permissions. Open it, check the numbers, adjust what needs adjusting, and send it when it's right.

Meet the crew

Four jobs your agents already know how to do

Give an agent the role, connect it once, and talk. Each one works inside your account with the permissions you grant and nothing more.

The estimator

Takeoff to bid, priced from your catalog

Reads the drawings with you, works the takeoff, then builds the bid from your products and labor rates. You review the numbers and send.

  • Claude Code
  • Claude
  • ChatGPT

bids_create · bid_line_items_create_many · products_list · labor_rates_list

The office

Inbox and calendar to CRM, every morning

Reads your email and calendar through its own connectors, turns real requests into leads and contacts, puts walkthroughs and due dates where they belong, and drafts the follow-up when a bid goes quiet. Keeps working after you close the laptop.

  • Grok Bot
  • Claude
  • ChatGPT

leads_create · contacts_update · lead_activities_create · draft_bid_followup

The controller

Change orders and pay apps, ready to approve

Drafts the change order when the GC adds scope, preps the pay application against the schedule of values, and summarizes the month.

  • ChatGPT
  • Claude
  • Grok Bot

change_orders_create · pay_apps_calculate · projects_summary

The field clerk

Crew notes to site log

With an agent that lives in a chat channel, your crew’s notes and photos become the daily log: personnel, work performed, photos attached, hours logged. OpenClaw runs on WhatsApp and Telegram natively.

  • OpenClaw
  • Hermes Agent

site_logs_create · documents_create_upload_url · time_entries_create

Every tool runs under a scope you see on the consent screen before you approve. Nothing reaches a customer unless the send scope was in what you approved, and you can revoke any agent in one click. How it stays safe

What a day looks like

What the agent said. What BuildWorkPro recorded.

An example day. The routine is our founder's; the jobs and names are illustrative.

  1. 6:40 AM
    Office agent, morning routine

    “Two GC emails overnight look like real work. Leads created for the storefront job and the curtain wall RFI, contacts linked. Walkthrough Thursday 9 AM is on your calendar.”

    • Recorded
    • 2 leads in the New stage, 2 contacts updated
    • Bid due date set on the lead
  2. 8:15 AM
    You, in Claude Code

    “New bid for the storefront job. Plans are in the folder: 12 openings, our standard hardware, usual labor rates.”

    • Recorded
    • Draft bid with 14 priced lines from your catalog
    • Waiting for your review
  3. 10:30 AM
    Office agent

    “The GC replied with a revised opening schedule. Filed it on the lead and logged the exchange. Your reply is drafted.”

    • Recorded
    • Attachment on the lead, activity logged
    • Nothing sent
  4. 1:15 PM
    You, in Claude on your phone

    “The office build-out wants two more doors. Draft the change order at our standard door rate.”

    • Recorded
    • Change order draft with 2 lines on the project
  5. 4:30 PM
    Office agent

    “The pay app for the office build-out is due Friday. Ran it against the schedule of values. The draft is ready for your review.”

    • Recorded
    • Pay application draft, calculated
  6. 6:00 PM
    Office agent, end of day

    “Today: 1 bid to review, 2 new leads, 1 change order draft, 1 pay app draft.”

    • Recorded
    • Nothing sent. Everything waiting for you.

Works with

Connect the agent you already use

Same address, same sign-in, same scopes for every one of them. Pick yours and follow the guide.

Claude

Available

Web, desktop, iPhone, Android

Add BuildWorkPro as a connector, sign in once, and talk. The bid is waiting in your account when you sit down.

Setup guide

Claude Code

Available

Terminal

One command adds the server. This is how our founder does takeoffs: the drawings in a folder, the bid priced from the catalog.

Setup guide

ChatGPT

Available

Plus, Team, Enterprise

Add a custom connector with the same sign-in and the same scopes.

Setup guide

Grok Bot

Available

macOS, Windows, iPhone · beta

xAI’s always-on agent on a cloud computer. Add BuildWorkPro as a custom MCP server; it reads your inbox and calendar through its own connectors and keeps the CRM current while you’re on site. Also covers the Grok app.

Setup guide

OpenClaw

Available

Self-hosted · WhatsApp, Telegram, Discord, Slack

Open-source agent that lives in your chat channels. The MCP servers you use in Claude Code work here unchanged.

Setup guide

Hermes Agent

Available

Self-hosted · Nous Research

Connects over MCP with sign-in built in. Configure it from the dashboard or the command line.

Setup guide

Any MCP client

Available

Cursor, custom agents, and more

The connection is an open standard. If your agent can add a custom MCP server, it can run BuildWorkPro.

Setup guide

Meta Muse

Testing

Web, iPhone, Android, WhatsApp

Launched September 8 with built-in connectors and the ability to connect services that have a public API using your credentials. We’re testing it against the BuildWorkPro API. No setup guide yet.

Developer docs

How it stays safe

An agent gets exactly the access you grant

You see exactly what it asks for

The consent screen lists every scope an agent requests before you approve. Read scopes only look. Write scopes create and update. Sending a bid, approving a pay app, or approving a change order are separate scopes, and the approval ones are flagged.

Approve or deny, then narrow

If an agent asks for more than you want it to have, deny and reconnect later. Agents like OpenClaw and Hermes can also be limited to a list of tools on their side, so a field agent never even sees the tools that send or approve.

You review, you send

An agent can draft the bid and the follow-up. A bid reaches a customer when you send it, or when the send scope was in the request you approved.

Revoke in one click

Settings → Authorized apps lists every agent you have connected. Revoke one and it loses access immediately.

Every record has an owner

Agents act under your login and your role. A lead, bid, or site log an agent creates is yours, follows your permissions, and is found and edited in the app like anything else.

Details for the technically curious: the MCP overview and the tools catalog list every tool and the scope it needs.

AI agents FAQ

Do I need to be technical to use this?

No. If you can send a text message, you can use it. You connect BuildWorkPro to your agent once, and from then on you just talk — describe the job, ask the question, request the change. There are no commands or prompts to memorize.

Which agents work with BuildWorkPro?

Claude (web, desktop, and phone), Claude Code, ChatGPT, Grok Bot and the Grok app, OpenClaw, and Hermes Agent, each with an install guide in our docs. The connection is an open standard, so any agent that can add a custom MCP server works too. We are testing Meta Muse.

Which agent should I start with?

The one you already pay for. If you use Claude or ChatGPT, connect that. If you want an agent that keeps working while you are on site, Grok Bot. If you want your crew to text an agent from the field, OpenClaw.

Do I need to run a server?

No. BuildWorkPro hosts the connection. You paste one address into your agent and sign in. OpenClaw and Hermes Agent are agents you host yourself, but the BuildWorkPro side needs nothing from you.

Can the agent send a bid to my customer without me?

Only if the send scope was in the request you approved on the consent screen; it is listed there by name. The workflow that works is the one our founder uses: the agent builds, you review in BuildWorkPro, and you decide what goes out.

Can my crew text it?

With an agent that lives in a chat channel, like OpenClaw on WhatsApp or Telegram, yes. The agent works under the account you connected. If your crew would rather use the app, unlimited users are included.

Does the agent connection cost extra?

No. Like everything else in BuildWorkPro, it is included in the $79/month plan — no AI add-on, no token bill, no higher tier. You do need your own account with the agent, and some of them, Grok Bot for one, require a paid plan on their side.

What is this technically?

For the curious: the connection uses the Model Context Protocol (MCP), an open standard that lets AI agents securely operate applications on your behalf. BuildWorkPro runs a live MCP server covering bids, projects, contacts, leads, site logs, change orders, pay applications, and more — the developer details are in our docs.

Your agents do the work. BuildWorkPro keeps the record.

Every feature, unlimited users, and the agent connection — all in the one $79/month plan. Try the whole thing free for 14 days.

$79/month flat · 14-day free trial · No credit card required