Tell it what happened.
The CRM is already updated.
Parsons CRM is 32 MCP tools with a console on top. Point the Claude or ChatGPT you already pay for at it — it logs the call, moves the deal, books the follow-up. Those tool calls are free and uncapped.
$15/mo flat, or $150/yr · no seats · your AI's tool calls free and uncapped · card collected at checkout · cancel any time
The software never broke. Thursday did.
You didn't stop using your last CRM because it lacked a feature. You stopped because logging one call meant opening a form built around objects and tabs, so you did it later, so you didn't, so the forecast went stale, so you stopped opening it. Adoption research puts CRM rollouts that miss their objective at roughly 55% (source) — and almost none of them fail because the software fell over.
| Contact | Stage | Last updated |
|---|---|---|
| Dana Reyes | proposal | 1 day ago |
| Marcus Bell | qualified | 3 days ago |
| Priya Nair | demo | 6 days ago |
| Tomas Alderman | negotiation | 9 days ago |
| Lena Vance | new | 12 days ago |
of a seller's working week goes into data entry — typing things they already know, into a form built for a database. source
is when the call gets logged, because logging it now means finding the record and filling the form. You will do it twice and then stop.
is the answer people give when you ask how out of date their pipeline is. At three weeks the number is a guess wearing a forecast's clothes.
Every CRM in this category answers that with a better form. We think the form is the problem. If the thing that updates the record is the assistant you are already talking to, the record is updated while you are still walking to the car.
Six CRM home pages. Zero prices.
We went and looked, in August 2026, at the six CRMs you are most likely weighing against this one. Not one of them prints what it costs on its own front page — and every one of them has a second meter waiting behind the first number. That is not sloppiness. It is what per-seat plus per-credit pricing forces you to do. Every competitor figure below is a published list price taken from the vendor’s own pricing page, checked August 2026.
Forecasting and sequences live on the tier above. HubSpot’s own footnote calls the one-time $1,500 Professional onboarding “required”. source
The headline number is the annual-commitment rate. The same Lite plan month to month is $24 — about 70% more — and LeadBooster, Projects, Campaigns and the rest are separate add-ons. source
The entry plan is capped at one user and 10,000 leads, with workflows switched off. The tier that has the automation is Growth, at $109 month to month, and phone lines and AI credits sit on top. Both figures fall on an annual commitment — $9 and $99. source
AI runs on credits. Another 5,000 workspace credits costs $85/mo on monthly billing, and the allowance renews each month on your billing date. source
Two things are actually going on in those slips. Per-seat compounds — every hire doesn't add a seat, it adds a seat across every tool you own. And the AI is the upsell: you already bought Claude or ChatGPT, and the CRM is selling it back to you by the credit. We do neither, because we sell neither.
The tools aren't an integration. They're the product.
Everyone else built a CRM for a mouse and bolted an MCP connector onto the side, so the connector can only reach the parts someone remembered to expose. Here, 32 tools are the whole feature list — and the console reads the same rows through the same data layer.
That asymmetry on the left isn't laziness. It is what happens when the tool layer is a translation of a UI-first product instead of the product itself. Ours went the other way round, which is also why the tool calls cost nothing — there is no separate agent surface to meter.
One sentence. Five tool calls. Nothing typed twice.
This is the whole daily mechanic. You say what happened; your AI picks the tools. Hover a call to see which part of the sentence produced it.
"Just got off a call with Dana at Northwind. Good demo, they asked for pricing. Move it to proposal, $18k, and remind me Tuesday."
"What's my weighted forecast on the consulting line this quarter?"
Note the last one. The read direction matters as much as the write: your AI can ask the CRM a question mid-conversation and answer you with a real number, because list_pipeline does the Σ(value × stage probability) server-side rather than handing back rows for the model to guess at.
A score you can do in your head.
Every contact carries a 0–100 score. Ours is five field checks and arithmetic — flip them and watch the number move. That is the entire model, and being able to reproduce it is the point.
→ {
"score": 85,
"rationale": "a short sentence naming
which of the five signals it found"
}
So "score this lead and tell me why" has a real answer — not a number you have to take on faith.
Here is the part nobody else will print: this is all it scores. There is no engagement tracking in this codebase — so none of the following exist, and we won't draw them on a landing page to look clever.
list_tasks orders your follow-ups soonest-due-first and search_contacts ranks matches highest-score-first. That is the real ordering, and it is the one we will describe.
Two businesses. One CRM. One forecast.
A consulting deal and a government bid are not the same motion, and forcing them through one pipeline is why most forecasts are fiction. Each business line gets its own stages — up to 12, ordered, each carrying its own probability. Click any deal to advance it and watch both the line and the combined total move.
A second pipeline here is one sentence — create_business_line, said out loud to your AI, on the $15 Base plan, with its own ordered stages and its own probabilities. Nothing to upgrade and no second workspace to buy. The default line ships as new → qualified → demo → proposal → negotiation at 10/25/50/70/85; you can leave it or replace it.
Σ(value × probability). Nothing more mysterious.
No model, no confidence score you can't audit. Each stage carries a probability; the weighted forecast is the sum of every open deal's value times the probability of the stage it is sitting in. Here it is on the shipped default ladder.
The probability rises every single stage — 10 → 25 → 50 → 70 → 85 — because a forecast where a deal gets less likely as it advances isn't a forecast. move_deal_stage carries the probability with the stage, so nobody has to remember to update a percentage field.
Win/loss reasons are a first-class field, not a custom property you have to buy a tier to create. Ask your AI why you lost last quarter and it has something real to read.
Paste the export. It sorts out the rest.
The reason people stay on a CRM they hate is the migration. Here it is one tool — import_contacts — and it is not a dumb loader: every row goes through three gates on the way in.
Skips any row whose email already exists in your book — and any row that repeats inside the same batch.
Matches the company name to an existing account, case-insensitively. If there is no match, it creates the account and links the contact.
"Cedar Labs" → Cedar Labs (created)
Every row is scored on the way in using the same five signals, so the book is ranked before you have looked at it once.
T. Alderman · no phone, no title → 40
Two honest numbers to hold onto. import_contacts takes up to 200 rows per call — 840 rows is five calls, and your AI makes them without being asked twice. And the plan holds 50,000 contacts: if your book is bigger than that, ask us to raise the ceiling from the billing console. We would rather you knew that before the import than during it.
There is a board. Nobody has to talk to a robot.
Chat-first does not mean chat-only. The console at crm.parsons.ai/app has four workspaces, and it reads the same rows your AI writes — there is no sync, no lag and no "connected app" to reconnect.
| Name | Account | Score |
|---|---|---|
| Dana Reyes | Northwind Retrofit | 85 |
| Marcus Bell | Bellcourt | 75 |
| Priya Nair | Cedar Labs | 60 |
| Lena Vance | Vance & Co | 100 |
| Account | Open value |
|---|---|
| Northwind Retrofit Ltd | $18,000 |
| Alderman Group | $52,000 |
| Bellcourt | $46,000 |
| Cedar Labs | $24,000 |
| Follow-up | Contact | Due |
|---|---|---|
| Send pricing for the retrofit | Dana Reyes | Tue 25 Aug |
| Chase the scoping call | Priya Nair | Thu 27 Aug |
| Renewal paperwork | Lena Vance | Mon 31 Aug |
That bottom strip is crm_stats — the same tool your AI calls when you ask how the quarter is going. The project switcher in the chrome is the other thing worth noticing: projects are unlimited and free, so a consultant with three books or an agency with a client each keeps them completely separate on the same $15.
There is a sales AI in the box.
The whole page so far assumes you already pay for a frontier model. If you don't, the console has a pinned chat with a model picker, running on the Parsons AI Gateway. Same tools, same rows — just our door instead of yours.
Be clear about what this costs
The $15 subscription is the access fee. It is not an AI bundle, and we are not going to pretend it is.
Tokens the built-in chat spends come out of your prepaid wallet, at provider cost × 1.5. You top it up; you set the cap; when the cap is hit it stops. There is no monthly credit allotment to expire, and nothing bills you after the fact.
Drive the CRM from your own Claude or ChatGPT instead and the AI costs you nothing here at all — those tool calls are free and uncapped.
One connection. Minted for you at sign-in.
There is nothing to configure and no API key to paste. Sign in, and your Parsons connection is created for you — add it once in Claude or ChatGPT and all 32 CRM actions are reachable through a single parsons_crm tool. That gateway shape is deliberate: MCP clients keep only the first 256 tools a server offers and silently drop the rest, so serving one action-parameterised tool per product is what keeps your whole book in reach. Hover any action to read what it actually does.
It carries the tools for every Parsons product you have access to, not just the CRM — and it is scoped to the project you are working in, so a client's book never appears in another client's session.
Get yours →The safety rail. Every destructive tool — delete_contact, delete_company, delete_deal, delete_task — refuses unless it is called with confirm: true. An agent cannot delete your book by misreading a sentence.
Deleting a company keeps its contacts and deals; it just unlinks them. And delete_business_line refuses outright on the default line or on any line that still has deals in it.
Four things it does not do.
A comparison that only flatters us reads as a lie, and you will find these out in week two anyway. Here they are in week zero.
| It does not | The truth, in full |
|---|---|
| Send email, send texts, or place calls | Zero reach tools. None of the 32 touch an inbox, a phone or a calendar. Parsons Email and Phone are real, but they are separate products, both in beta, with their own pricing: a US local number is $3.99/mo ($4.99 toll-free), voice is $0.03/minute and outbound SMS is $0.02/segment from the prepaid wallet. That is not included in the $15, and we are not going to draw a sent email on this page. |
| Watch for stale deals on its own | No cron, no triggers, no webhooks — nothing in the CRM wakes up without you. If you want a Monday-morning pipeline review, your Claude schedules it in your own AI client, or you wire Parsons Scheduler, which is a separate live product. Any page that shows a CRM chasing your deals overnight is describing the customer's AI, not the CRM. |
| Export your book | There is no export tool and no export button. Worse, you can't reconstruct one through the agent either: list_contacts has no cursor or offset, so it reads the newest 100 per call and no further. We are not going to sell you a date for one. If walking out with a CSV on demand is a condition of buying, this is the row that should stop you — and we would rather it stopped you today than in month six. |
| Behave like a data warehouse | Search is a substring scan across name, email and title. Contacts read newest-first, capped per call. That is completely fine at a thousand contacts and one seller; it is not fine at fifty thousand and a twenty-rep floor. If that is you, we are the wrong shape and you should buy the expensive one. |
Also worth saying plainly: there are no sequences, no dialer, no enrichment, no email sync, no territory rules and no per-rep quota reporting. Every one of those is a real reason to buy something else.
What five people costs.
Seats are the whole argument. Drag the slider: five lines climb and one does not, because seats are not our unit — the subscription covers the account, not the head count.
Every competitor row is that vendor’s published list price for month-to-month billing, checked August 2026. Each links to the page it came from. Annual commitments are cheaper at all five — HubSpot $90, Close $99, Attio $79, Pipedrive $59, folk $48, each still per seat, per month — and each locks you in for a year. Ours is $150/yr for the whole account: two months free, no seat multiplier, and the $15 monthly price is unchanged if you skip it.
$15. That's the whole menu.
Monthly or yearly — yearly is exactly two months free, and nothing else about the plan changes. No per-seat multiplier, no enterprise tier to be routed into, no free tier. For the CRM, exactly one number is worth memorising: 50,000 contacts.
- 50,000 contacts — accounts, deals, activities and follow-ups are not separately capped
- All 32 MCP tools, and every tool call is free and uncapped
- Business lines are not capped — each carries up to 12 ordered stages with their own probabilities
- The console — Pipeline, Contacts, Accounts, Tasks — plus the built-in sales chat
- Every other Parsons product on the same $15 — one subscription covers them all
products, one subscription
The same $15 covers every live Parsons product — CRM, Money, Tracker, Journal, Auth, the MCP Hub, Scheduler and the rest. One login, one bill.
projects, at no extra cost
A project per business, or one for work and one for personal. Separate data, separate members, separate spend caps, at no extra cost — the subscription covers the account, not the workspace count. (Contacts are pooled account-wide — see above.)
days, free, then $15
The trial is a real subscription with the first 14 days free, on your first subscription only. Checkout collects a card. Cancel any time before it renews and you are not charged.
"Is a CRM worth $15 a month?"
That is the right question to ask, so here is the arithmetic rather than an adjective. $15 is roughly a quarter of what one seat of the cheapest CRM on that slider costs — and a seat is where their bill starts, not where it ends. What you are weighing is one seat of one product against a whole account, thirteen products and every project you care to make.
And if you want it cheaper still, yearly is $150 — two months free — with the same cancel-any-time monthly price waiting if you would rather not commit. The CRM's own usage is the clearest example of the shape: we count your contacts, and we chose not to bill them. The ceiling is a stop you can ask us to raise, not a surprise line item.
$15/mo or $150/yr · yearly is two months free
14-day free trial on your first subscription · card collected at checkout · cancel any time
no per-seat pricing · no enterprise tier · no free tier
The questions you were going to email us.
Do I need Claude or ChatGPT to use this?
No. There is a full four-workspace console and a sales chat built in. But the CRM is at its best when your own AI drives it — that is what it was designed around, and those tool calls are free and uncapped.
Does the AI cost extra?
Tool calls from your own Claude or ChatGPT to the CRM: never. tool_calls is metered in the usage registry with no plan cap. The built-in sales chat is different — it draws on prepaid wallet credit at provider cost × 1.5, and you set the cap.
What happens when I hit 50,000 contacts?
The next write stops, and the billing console has an ask-us-to-raise-it path if your book is genuinely that big. No overage bill, ever. One thing to know: contacts are counted across every project your account owns, not per project, so extra projects don't multiply the allowance.
Is there a free plan?
No. There is a 14-day free trial on your first subscription, then $15/mo — or $150/yr if you would rather pay yearly, which is two months free. Checkout collects a card so the subscription can continue if you keep it — cancel any time before it renews and you are not charged. Cancelling and resubscribing later does not start a second trial.
Can I get my data out?
Not today. There is no export tool and no export button, and because list_contacts has no cursor, your agent can't page through the whole book to build one either. We are not going to promise you a date — we would rather tell you now than have you find out at the wrong moment. You can delete everything for real: purging a project wipes its rows.
Can it send the follow-up email it just booked?
No. The CRM has no reach tools at all. Parsons Email and Phone are separate products, both in beta, with their own pricing — $3.99/mo for a US local number, $0.03/voice minute, $0.02 per outbound SMS segment.
Can it chase deals that have gone quiet, by itself?
Not by itself — it has no cron, no triggers and no webhooks. Ask your Claude to review the pipeline and it will, and your AI client can put that on a schedule. Parsons Scheduler (a separate live product, same subscription) can also drive it.
I run two businesses. Do I pay twice?
No. Separate projects, separate data, separate members and spend caps, same $15. Projects are unlimited and free. The one caveat is the contact ceiling, which is pooled across the account rather than per project.
Can my AI delete something by mistake?
Not by misreading you. Every destructive tool refuses unless it is called with confirm: true, which is a deliberate second step after you say yes. Deleting an account keeps its contacts and deals and only unlinks them.
Say what happened. The rest is already done.
$15/mo, or $150/yr for two months free. No seats, no expiring AI credits, no overage bill. 14-day free trial, card collected at checkout, cancel any time.