Know what your quarter is worth, not what you hope it is worth

A kanban pipeline with stages, probability and a weighted forecast. Every deal links optionally to a contact and a company, and every stage change is written to the audit log.

Want to see it first? Book a demo

Weighted forecast built in
Stage changes audited
Same actions over MCP

app.salescrew.io/deals

Pipeline with weighted forecast and stage breakdown

The short answer

SalesCrew gives every deal a stage, a probability and a weighted value, so the forecast is a sum of the board rather than a separate spreadsheet.
Contact and company links on a deal are both optional, which means a marketplace bid or a referral can be a deal on day one without inventing a company record.
When an agent moves a deal to a new stage, that move waits in the approval queue instead of changing the number behind your back.

What it does

The Deals board is a kanban view of your pipeline. Each column is a stage you named, each stage carries a default probability, and each card shows the deal value and the weighted value that probability produces. Drag a card to another column and the weighted total at the top of the board changes with it. Pipeline stages are yours to define: the same board serves a fixed-fee build shop and a monthly retainer agency because the stage list, and the probability on each stage, is configuration rather than code.

Open a deal and you get the record, not a link out to five other tabs. Tasks with owners and due dates, meetings, notes, files and a timeline of everything that has happened sit on the deal itself. The timeline pulls from the same database as the inbox, so an email thread that moved the deal appears in its history without an integration step. Contacts and companies each have the same shape, and a deal can point at one, both or neither.

Every button on that board is also a tool on the MCP server. deal_create opens a deal, deal_update_stage moves it, deal_get reads one back with its links, and pipeline_query returns the board as data an agent can reason over. Stage definitions themselves are readable and writable with pipeline_stage_list and pipeline_stage_upsert. stalled_deals returns the deals that have not moved, and task_create puts the follow-up on someone's list. The effect is identical either way: a human dragging a card and an agent calling the tool write the same row and leave the same audit entry, with the actor recorded as a user, a token or an agent.

Lead scoring writes a score with its reasons and surfaces a next action on your top-scoring records. Record summary condenses a long timeline into one paragraph, so you can read a deal's history in the time it takes to open it.

What is different here

Each of these comes from one of the five ideas the product is built on.

The board is also an API

Agent-operable by design: every action on the deal board exists as an MCP tool with the same effect and the same permissions. Nothing on this page is reachable only by mouse.

Agents propose, you dispose

Human in the loop: stage changes, merges and deletes default to review. An agent's move lands in the approval queue with its evidence, and you approve, edit or reject it.

Links are optional, on purpose

A deal can exist without a company or a contact. Upwork bids, referrals and inbound forms become pipeline immediately instead of waiting for someone to tidy the data.

One database behind the board

The whole funnel in one place: the reply that moved the deal, the cadence it came from and the ad spend that sourced it all live in the same Postgres project you own.

How you get to a working board

The same four steps apply to every module. Nothing runs on auto until you move it there.

1
Connect
Import contacts and companies, or connect a mailbox so threads attach themselves to the right records from the start.
2
Configure
Name your stages, set a probability on each one, and decide which fields your team actually fills in.
3
Run in review
Turn on scoring and agent suggestions in review mode. Every stage change an agent proposes waits in the queue with its reasoning.
4
Move to auto
After a clean week, promote the low-risk classes — logging, tagging, task creation — to auto. Stage changes can stay in review as long as you like.

In which tier

The pipeline is not an upsell. It is on every plan, with unlimited seats.

Tier	Price	What you get on the pipeline
Core	$299/mo	Full deals board, stages, probability, weighted forecast, tasks, meetings, notes, files, timeline. MCP tools included.
Local	With the phone and SMS pack	Core, plus phone and appointment features that write back to the deal.
Outbound	$699/mo	Core, plus in-product agents that can propose stage changes and next actions inside the approval queue.
Agency	$1,200/mo	Outbound, plus marketplace deals and the marketing module beside the same board.

Seats are unlimited on every tier.

Bring your records in without a mess

You export contacts, companies, deals, notes and tasks from your old CRM as CSV. Contacts go into the data bank through the CSV import: map the columns, tag the batch, and rows match existing contacts on email, filling gaps without overwriting. Companies and deals follow through the MCP tools company_create and deal_create, and the import wizard walks the whole export through the same path. Duplicate detection and merge clean up any pairs that slipped through.

Migration help is included on every tier, and a standard import completes within 7 business days. Custom fields, AI-prompt fields, quotas with a leaderboard, and forecast categories shape the board around how your team sells. You can take a full export of everything at any time.

Questions

We already have a CRM with a pipeline. Why move?: Two structural reasons. An agent can operate this board through 140+ MCP tools with the same permissions a person has. And your pipeline sits in a Postgres database that belongs to you, not a shared tenant. Migration help is included on every tier, and a standard import completes within 7 business days.
Can I define my own stages and probabilities?: Yes. Stages, their order and the probability on each are configuration, editable in the app and through pipeline_stage_upsert. The weighted forecast recalculates from whatever you set.
Does a deal have to belong to a company?: No. Contact and company links are both optional. That is deliberate: marketplace bids often have no company name available, and forcing one produces junk records.
What stops an agent from mass-moving deals?: Stage changes, merges and deletes default to review, so they queue for a person. Guardrails and a kill switch that stops every agent at once sit on the Agents page, and every tool call is written to the audit log with its actor.
What is on a record?: The standard contact, company and deal fields plus notes, tasks, meetings, files and a timeline. Custom fields hold anything your team tracks, and AI-prompt fields fill themselves from a prompt you write.
How do I get my data out?: Your instance is its own Postgres project. Export is a full database dump plus CSV everywhere, not a partial export of selected objects.

Put a real forecast on one board

Import your contacts, name your stages, and run the first week with every agent action in review.