The FUB-native AI you also use on the go in Claude or ChatGPT.
189 tools, scoped to your role. Sign in with your FUB email. Update your pipeline from anywhere.
OAuth with your FUB email · no API keys to paste · tools scoped to your role · works in Claude, ChatGPT, Cursor and more
You're not at a desk. Your CRM should know that.
The Tuesday between a 10am open house and a 2pm showing. Coffee at 11. Drive across town at 12. Your laptop is closed. Your phone is in your hand. The lead that came in at 10:14 is still waiting.
This is the moment every other AI CRM tool fails — they make you sit down. Ace doesn't. Paste a URL into Claude.ai or ChatGPT, sign in with your FUB email, and the AI that runs your phone now also runs your CRM. Open-house parking lot. Between showings. In the car. Anywhere your phone goes, Ace goes.
That's the wedge. No competitor has it.
One Ace. Four places it lives.
Pick the surface that matches where you are. The AI is the same — same FUB context, same memory, same tools.
Claude Connector
Use Ace from claude.ai — desktop or mobile browser. The fastest install, the most polished tool-use UX.
https://followupace.com/mcp
- Open claude.ai → Settings → Integrations → Add
- Paste the URL above
- Sign in with your FUB email
- Enter the verification code we email you
ChatGPT Connector
Use Ace from ChatGPT. Same OAuth. Same tools. Implements the polymorphic search + fetch ChatGPT requires.
https://followupace.com/mcp?tools=full
- Open ChatGPT → Settings → Security and login and enable Developer mode
- Open ChatGPT Plugins and select the plus button
- Paste the URL above
- Sign in with your FUB email
The iPhone app
The one surface built for a phone in a driveway. Opens on the person to work next with the draft already written; the iPhone's Messages composer opens pre-filled and you tap Send — from your own number — and the send is logged back to the Follow Up Boss timeline.
- Download Follow Up Ace free on the App Store
- Enter your FUB email; we email a 6-digit code
- Your queue is already ranked and waiting
iOS 15.0+, iPhone & iPad. iPhone only — Android and desktop use the browser Companion.
About the iPhone app →Mobile voice (in FUB embed)
Open FUB on your phone. Tap the Ace embed. Hold the mic. Sub-2-second WebRTC speech-to-speech via ElevenLabs Flash v2.5. Mobile-bulletproof on iOS Safari iframes.
- Open Follow Up Boss on your phone
- Tap the Ace embed (it's already inside FUB)
- Hold the mic and ask
Read-only by design. Voice answers, summarizes, and prepares — you tap Send.
Power-user footnote: prefer a native install? Claude Desktop via DXT is a different install path, not a bigger toolset — the extension proxies this same production Connector, so it serves the same role-scoped surface (up to 189 tools) that claude.ai and ChatGPT get. Pick it if you want a one-click desktop install; pick the web Connector if you want zero install.
Nobody is shown a tool they can’t run.
The Connector used to advertise one identical toolbox to everyone who signed in. An agent seat could see the team leaderboard tool, call it, eat a permission error, and burn a turn guessing why. That’s fixed. The tool list is now built from your Follow Up Boss sign-in, so what the AI can see is exactly what your role can actually do.
Team lead / admin (with a seat) — 189 tools
- Everything an agent gets, plus the team-wide layer
- Production and response time by agent, conversion funnel, source ROI
- Back Office commission ledger, You vs Market, Revenue Guard summary
- Ace Smart Lists: plan the account’s lists, then deploy them into FUB
- 182 without the Ace Trove add-on — the Trove-gated tools drop out
Individual agent seat — 170 tools
- The full Follow Up Boss CRM against your own book
- Your NOW queue, your daily brief, your alerts, your memory
- Your commissions — not the team’s ledger
- Your own notification preferences, readable and writable
- Cross-agent governance tools are simply not listed
A seatless admin — someone who administers the account but holds no Follow Up Boss seat — sees 161 tools, because the “my book / my alerts / my commissions” tools have nothing to answer for them.
Scoping is not a display trick. The tool list is filtered when your AI asks what’s available, and re-checked at call time, so a tool that was never listed still can’t be invoked by a model that guessed the name. A denial comes back as a structured message telling the AI where to route instead — not a raw permission error it has to interpret.
189 tools on the Connector surface — memory, scoring, seller intent and compliance, not just CRUD.
The full advertised surface, by category. Your own list is a subset of this, scoped to your role — ask your AI to list its tools after you connect and it will tell you exactly what you got.
13 new tools that reach past Follow Up Boss into the rest of Ace.
Anyone can wrap the Follow Up Boss REST API. What you can’t wrap is the part of your business that doesn’t live in Follow Up Boss. These tools put the rest of the Ace platform into the same conversation — so “how much did I earn on closings this quarter” and “who filled out my open-house form” are answerable in the same breath as “who haven’t I called.”
You vs Market is peer benchmarking with a k-anonymity floor of 10 — a comparison only computes when at least ten accounts are in the cohort, so it can never expose a single competitor’s numbers. Revenue Guard dollars are modelled from your account’s own median commission and are labelled as estimates. Back Office credits a multi-agent deal in full to the primary agent and flags it, because Follow Up Boss carries no split ratio and we won’t invent one. Smart Lists deploy and You vs Market, Revenue Guard and the Back Office ledger require the Ace Trove add-on or team-lead access as noted.
Every answer links back to the record. Every write gets checked.
The answer ends where the work starts.
An AI that says “Sarah Chen hasn’t been contacted in 18 days” has handed you a second job: go find Sarah. Now the answer carries the link — straight to that person’s Follow Up Boss record, and to the Ace screen that shows the reasoning. Which Ace screen depends on who you are: an agent gets Companion links to their own queue, an admin gets the admin dashboard.
app.followupboss.com/2/people/view/9976
An unverified write is never reported as a success.
- Read-back verification. After a person or deal is created, edited, deleted or merged, Ace reads the record back and compares it against what you asked for. Three verdicts — verified, unverified, contradicted — and only verified is narrated as done.
- Preview before account-wide writes. Deploying Ace Smart Lists touches every user in the account, so the tool refuses to write without an explicit confirmation and returns the exact create/update/delete plan first.
- Bulk operations are fenced. In the developer package, bulk tools cap at 100 records per call, require explicit confirmation above 10, return a preview instead of writing when confirmation is missing, and abort the loop after five consecutive upstream failures rather than grinding through a broken list.
- Outbound sends stay agent-reviewed. A queued message is drafted, held, and only sent when you confirm it.
Would create 4 lists, update 2, remove 1 duplicate. Nothing was written.
Every tool is annotated as read-only or state-changing, and destructive tools are marked destructive. Your AI host reads those annotations: safe reads can be auto-approved so a lookup doesn’t interrupt you in a driveway, while anything that deletes or sends stops and asks. That’s the difference between an integration you can use one-handed and one you have to babysit.
One connector. Whatever AI you already use.
MCP is an open protocol, so the same endpoint answers to every client that speaks it. You sign in with your Follow Up Boss email over OAuth — there is no API key to paste into a config file, and none to leak, rotate, or store on a laptop.
- Claude — web, desktop, mobile, and Claude Code.
- ChatGPT — as a connector, and in Deep Research via the search/fetch pair ChatGPT requires.
- Cursor, Copilot, Gemini, LibreChat — and anything else that speaks MCP.
- Claude Desktop via the DXT extension — a one-click install path that proxies this same connector, so it is the same role-scoped surface, not a bigger one.
You could wrap the Follow Up Boss API yourself. You couldn’t build the half that matters.
Follow Up Boss does not ship an MCP server. The third-party ones that exist are local wrappers over the public REST API — you paste your API key into a config file, run a process on your laptop, and the AI gets a conversational front-end to the API. That is a genuinely useful weekend project, and if all you want is “create a note by voice,” build it.
The line worth understanding is what a REST wrapper structurally cannot reach. Everything below needs infrastructure that runs whether or not your laptop is open, and history that had to be accumulating before you asked the question.
What an API-key wrapper gives you
- Whatever the REST API returns, today, on request
- One shared key with one permission level — no per-person roles
- A key sitting in a config file on a machine you control
- No memory between conversations beyond what you paste in
- No history it didn’t just fetch
What needs multi-tenant infrastructure
- Scored contacts. Win Score and Churn Risk come from models trained on accumulated outcome history and calibrated against what actually closed — not from a field the API hands back.
- Peer benchmarks. Comparing your numbers to the market needs other accounts’ numbers, anonymized at a k-anonymity floor of 10. One tenant cannot benchmark against itself.
- Deal-activity attribution. Knowing whether a call was pushing a closing or working a lead requires having watched the webhook stream while it happened.
- Seller intent. Tenure, estimated equity and area signals drawn from 30+ third-party data sources, refreshed on a schedule.
- Continuous scoring. Contacts are re-scored on a cadence and written into Follow Up Boss fields, so the answer is ready before you ask.
An API key buys you access. It doesn’t buy you a warehouse of your own webhook history, a scoring model calibrated on outcomes, or a cohort of anonymized peers to be measured against. Estimated home equity is modelled and always labelled “(est.)” — never appraisal-grade. Move signals come only from what an agent recorded in Follow Up Boss; we don’t buy third-party life-event data. Those are the boundaries of what the intelligence layer actually is, and we’d rather you know them going in.
Your day-in-progress follows you.
Draft a follow-up in Claude on your commute. Confirm it from inside Follow Up Boss when you sit down. Your pending actions, your morning brief, and your day-in-progress follow you across phone, desktop, and voice.
Pending: Send check-in text to Sarah Chen
"Hey Sarah, just saw 2847 Oak St had a price drop today — thought you'd want to know. Want to swing by this weekend?"
Confirm send Cancel
The Session Dock is a UI pattern in the embed — pending actions and the brief carry across phone, desktop, and voice. The embed chat AI does not yet ingest a session block at the prompt level (i.e., "the AI inside FUB knows your ChatGPT history" is not yet shipped). What's real is the dock. We won't oversell.
Per-Pro-user OAuth. Permissions inherit from FUB.
Every agent on your team signs in with their own FUB email — no API key is ever pasted, stored, or shared. Ace inherits FUB role-based permissions transparently: admin OAuth gets admin tool scope, agent OAuth gets agent scope, and the scope decides both what is listed and what may be called.
- Short-lived, revocable access. OAuth credentials are scoped per user, expire, and can be invalidated immediately.
- Revocable on demand. Bump
mcpSessionVersionon the user document — every previously issued token invalidates immediately. - Email verification required. The OAuth flow mails a code to the FUB inbox — proves the user owns that mailbox.
- No password sharing. Each agent installs the Connector once with their own credentials. The team-lead doesn't see anyone else's tokens.
The questions agents ask first.
What's a Connector?
It's how Claude.ai and ChatGPT plug into external tools. You paste one URL, sign in once with the same email you use for that tool — and the AI on those platforms can now read and act on that tool's data. Ace is a Connector for Follow Up Boss. You don't need to know any more than that.
Do I need to be technical?
No. Paste a URL into a settings panel. Sign in with the FUB email you already have. Type the verification code from your inbox. You're done in under 60 seconds. Most agents finish on a phone before their coffee is ready.
What does Ace see when I connect?
Exactly what your FUB user account can see. Ace inherits your role-based permissions. Admins can act across the team; agents are scoped to their own contacts. Emails and phone numbers are tokenized as opaque placeholders before they're sent to the LLM, then resolved at tool-call time — your contact data never leaves your account in cleartext.
How many tools do I actually get?
189 are advertised on the Connector, and your list is a subset scoped to your role: the full 189 for an admin who also holds a Follow Up Boss seat, 170 for an individual agent seat, 161 for an admin with no seat. Without the Ace Trove add-on, the Trove-gated tools drop out of those lists (182 / 166 / 157). Don't take our word for it — after you connect, ask your AI to list its tools. That answer is authoritative.
Does Claude Desktop give me more tools than the web Connector?
No. The Claude Desktop DXT extension proxies this same production Connector, so it serves the same role-scoped surface. It's a different install path — one click, native to the desktop app — not a different or larger toolset. There is a separate stdio developer package with a larger registry (270 tools), but that's a build target for developers, not the endpoint a team connects to.
Can it change or delete things without asking me?
Every tool is annotated as read-only or state-changing, and destructive tools are flagged as destructive, so your AI host knows to stop and ask before anything irreversible. Writes that create or edit a person or deal are read back and compared to what you asked for — an unverified write is never narrated as a success. Deploying Smart Lists across the account refuses to write at all until you've seen the plan and explicitly confirmed. Outbound client messages are drafted and held for your review, not auto-sent.
Which AI assistants does it work with?
Claude on web, desktop, mobile and Claude Code; ChatGPT as a connector and in Deep Research; Cursor, Copilot, Gemini, LibreChat, and anything else that speaks MCP. You sign in with your Follow Up Boss email over OAuth — there's no API key to paste anywhere.
Can I revoke access?
Yes — instantly. From the admin dashboard, bump mcpSessionVersion on the user (or hit "Revoke MCP access" in the user menu) and every previously issued JWT for that account invalidates immediately. We surface a one-click revoke for any compromised laptop or off-boarded agent.
Do I need a paid plan to use Connectors?
Connectors require an Ace Pro seat ($55/mo). The signup itself is free — admin signs up at followupace.com, connects FUB once, and your whole team gets free Ace accounts automatically. To enable the Connector flow on your seat, ask your admin to upgrade you to Pro (takes a minute, no contracts).
We build transparency in — so you get better at AI, not dependent on it.
Wherever Ace works — Claude, ChatGPT, or inside Follow Up Boss — it shows its work. In the app, every recommended move surfaces the exact, full prompt it runs, so you learn to direct AI in your own words; client-facing messages are agent-reviewed before they send; and the whole process is regression-tested in CI with a public evolution log. The agents who pull ahead aren’t the ones who lean on AI — they’re the ones who learn to direct it.
Connect once. Use Ace from anywhere.
Paste a URL, sign in with your FUB email, get back to work. Sixty seconds, max.