// hubspot-mcp/server.ts
import { Server } from "@modelcontextprotocol/sdk/server";
import { hubspot } from "./client";
const server = new Server({
name: "hubspot-mcp",
version: "0.1.0",
});
server.tool("contacts.search", async ({ query }) => {
return hubspot.contacts.search({ q: query });
});
server.tool("deals.update", async ({ id, patch }) => {
return hubspot.deals.update(id, patch);
});
server.listen();
Live · AI agents for B2B SaaS
Ship AI agents and MCP integrations that move real work.
Production automations for B2B SaaS teams. n8n, Zapier, and MCP integrations — wired into HubSpot, Linear, Notion, Slack, and the rest of your stack via Workato, Zoho MCP, and native LLM tool-calling. Audit in a week. Sprint in two. Built and operated by one engineer who ships.
nextflow ~ agent runtime
live
live
/
v0.4.2
/
12 prod agents
/
last ship 2h ago
/
99.8% uptime · 30d
Real code, shipped into real stacks.
Three patterns that show up in nearly every engagement. Each card is an actual artifact — the kind of thing that goes into your repo on day one.
// churn-handoff.json
{
"name": "churn → CS handoff",
"trigger": "stripe.subscription.deleted",
"nodes": [
{
"id": "find_owner",
"tool": "hubspot.contacts.lookup",
"by": "email"
},
{
"id": "create_ticket",
"tool": "linear.issue.create",
"team": "CS", "priority": 2
},
{
"id": "notify_owner",
"tool": "slack.chat.postMessage"
}
]
}# churn → CS · Workato recipe
recipe "churn handoff" do
trigger Stripe::SubscriptionDeleted
step "find owner",
HubSpot::Contacts.lookup(
email: trigger.customer.email
)
step "create case",
Zendesk::Tickets.create(
subject: "Churn followup",
assignee_id: lookup.owner_id,
priority: "normal"
)
step "ping owner",
Slack::DM.send(to: lookup.slack_id)
endThree productized shipping missions.
Fixed price, fixed scope, fixed timeline. No retainers required to start. Most teams begin with the audit; expanding into a sprint or a setup is the common path.
$499
AI Automation Audit
A surgical look at where AI and automation can replace manual work in your stack.
- 2 discovery calls with your ops/RevOps lead
- Full stack map — tools, gaps, time-leak ranking
- 30-day AI-automation roadmap, prioritized by ROI
- One quick-win blueprint, ready to build
- Async Loom walkthrough on delivery
$1,750
Workflow Sprint
2–3 production workflows in n8n or Zapier, with one AI-powered step.
- 2–3 production-ready workflows (n8n or Zapier)
- One AI step — classification, summarization, routing, drafting
- Full documentation + Loom walkthrough
- 14-day post-ship support
- Telemetry hooks for monitoring
$3,500
AI Agent / MCP Setup
One AI agent or MCP integration set up on your stack.
- One AI agent or MCP integration
- Existing MCP servers (Zoho, GitHub), MCP-native platforms (Workato), or LLM tool-calling
- Prompts + tool definitions + error handling + observability
- 2–3 tool integrations (HubSpot/Linear/Notion/Slack/Intercom/Stripe)
- 30-day post-ship support + one iteration round
- Custom MCP server build · add-on from $1,500
Five stages, no surprises.
Same shape every time. Audit fits inside a calendar week. Sprints and setups extend stages 3 and 4.
-
01
Brief
30-min call. What's bleeding time, what's already automated, what's politically off-limits.
~30 min -
02
Map
Async stack walkthrough. Read-only access to your tools. Identify the 3–5 highest-leverage interventions.
~2 days -
03
Build
Build in a private staging env. Daily Loom updates — you see progress every 24 hours.
1–3 weeks -
04
Ship
Production deploy. Documentation handoff. Loom walkthrough. You own everything — code, prompts, recipes.
~1 day -
05
Tune
Post-ship support window. Tweaks, bug fixes, edge cases. Optional move into Care retainer.
14–30 days
What we connect.
Not a logo wall. Each tool below describes how it actually plays in a NextFlow build — what we trigger from, what we write to, what we orchestrate.
HubSpot
Source of truth for contacts and deals. Trigger on lifecycle stage. Update from agent outputs.
Salesforce
Bulk-record sync, lead routing, opportunity hygiene. SOQL for the things native connectors miss.
Linear
Auto-created issues from churn signals, support escalations, or agent-detected anomalies.
Notion
Knowledge-base reads for RAG, doc-driven workflow triggers, structured-database writes.
Slack
Owner alerts, agent-handoff DMs, daily summaries, slash-command triggers for ad-hoc runs.
Intercom
First-line CS automation, ticket routing, agent-drafted replies for review-and-send flows.
Stripe
Churn detection, MRR-change handoffs, dunning automation, subscription-event triggers.
OpenAI
GPT-class drafting, classification, structured extraction, native function-call tool use.
Anthropic
Claude for long-context reasoning, agent loops, MCP-native tool use, careful drafting.
n8n
Self-hosted workflow runtime — durable, observable, version-controlled. Default for sprints.
Zapier
Fastest path to a working flow when n8n is overkill. Five-plus years of muscle memory.
MCP
Setup of MCP servers — official (Zoho, GitHub, others), MCP-native platforms (Workato), or custom builds when you really need one.
What's been shipped in the lab.
Lab builds — projects shipped to prove patterns, not paid client engagements (yet). Tap a row to expand. Live walkthroughs available on request — much faster than a static page.
FR-01
·
·
v0.1 · lab
CRM Sync Dashboard
Zoho ↔ HubSpot field-mapping with webhook-driven sync and telemetry. Currently a controlled-data demo; real OAuth + live API rebuild is the next ship.
FR-02
·
·
shipped
AI-Driven Support Chatbot
First-line CS automation on Crisp and Zopim. ChatGPT + Gemini for intent detection, drafting, and confident escalation to humans when the model isn't sure.
FR-03
·
·
shipped
No-Code API Integration Portal
A Bubble.io self-serve UI that lets non-developers configure API automations — Bitrix24 to comms apps, no engineer required. Reduced deployment time substantially.
Things people ask.
Where are you based and what timezone do you work in?
Based in India (IST). Day-to-day operating window aligned to 9 AM EST onwards — meaningful overlap with US East Coast and full overlap with US Pacific evening. UK clients catch me through their morning. Async-first; calls fit your calendar, not mine.
How does billing work?
Audit is paid up front. Sprints and setups split 50/50 — half on signing, half on delivery. Care retainer is monthly, 3-month minimum then monthly cancel. Payment methods: Wise, AirTM, PayPal, or bank wire — whatever your finance team prefers. We confirm the right rail at signing. Invoiced in USD.
NDAs? IP? Who owns the code?
NDAs welcome — I sign yours, or use a clean mutual NDA on request. You own the work product end-to-end: code, prompts, recipes, documentation. I keep no proprietary IP claim on client builds. Open-source dependencies stay under their licenses.
What's explicitly out of scope?
I don't build production SaaS apps, mobile apps, or marketing sites. I don't run paid ads or manage social. I don't take on roles that require being on-call 24/7. Anything outside that — automation, AI agents, integrations, internal tooling — is fair game.
Can I pause or cancel the Care retainer?
3-month minimum to start (so we both have runway to do real work). After that, you can pause for up to 60 days no questions asked, or cancel with 30 days notice. Unused hours don't roll forward — the retainer buys ongoing capacity, not a banking arrangement.
Why one engineer instead of an agency?
Because agencies introduce account managers and PMs between you and the person actually writing the code. With NextFlow, the engineer takes the brief, designs the system, ships the build, and answers the support thread. Twelve-plus years of operating experience — the kind of background that sees the political angles, not just the technical ones.
Send me a brief.
Fastest path: tell me your stack and the workflow that's eating hours. Reply within 1 business day with a recommendation — audit, sprint, setup, or "not the right fit."