Agent-written account briefs

Two MCP servers, no integration build, and a researched brief on the record before the call.

Every other workflow on this site is a pipeline: a trigger, a set of steps, a schedule and something to maintain. This one is not, and that is the point of including it. An AI client connected to two MCP servers, one for data and one for the CRM, can do the research-and-write-a-brief job without anybody building the integration between them first.

That is a genuinely new capability rather than a faster version of an old one. The job it replaces is not a pipeline, it is the twenty minutes a rep spends before a call assembling context from four tabs, or the half-built automation somebody wrote to do it badly. It does not replace a scheduled sync, and this page is careful about that line.

What this is and is not

This is on-demand research with a write-back, driven by a person or a task rather than by a record event. It is not a substitute for the CRM data layer, which is a scheduled loop over thousands of records and should stay one. An agent is the right tool for one account at a time and the wrong tool for ten thousand.

The pipeline, step by step

4 steps. The ones in blue are the data layer; everything else is plumbing you already own or can stand up in an afternoon.

01

Add the Reverse Contact MCP server

Point the client at the MCP endpoint. OAuth 2.1 where the client supports it, otherwise an Authorization header with your API key. Thirty-seven tools arrive by their own count, 36 of them in its tables: enrich and search for people and companies, the live variants, the search filter catalogues, asynchronous email and phone lookups that return a task id to poll, Lists, Bulk, tasks and goals, and free checks on a profile, your usage and a running task. A company domain now goes straight in through enrich_company, which it did not before November 2026.

37 toolsone URL
↓
02

Add the CRM's own connector

HubSpot publishes a connector for Claude that sits at 184,000 installs on its marketplace, which makes it one of the most-installed apps HubSpot lists. It reads and writes CRM records under your own permissions.

184Kinstalls, HubSpot marketplace
↓
03

Write the prompt once, reuse it

Name the account, say which fields to pull, say what the brief should contain, and say explicitly what to do when the data is missing. The last instruction is the one that decides whether this is useful or dangerous.

1 promptreused
↓
04

The agent writes the brief onto the record

A note on the company record, so it is where the rep already looks rather than in a chat log. Optionally a Slack message to the owner, which is the same restraint rule as everywhere else: one channel, one trigger, one owner.

0 linesof glue code

Build it with an agent

This workflow ships as a skill file: the API surface of every tool in the chain, the build order, the limits that break it, and how to check it worked. Point an agent at it and it has what it needs without reading 3 sets of vendor docs first.

mcp-account-brief/SKILL.md agent skill
---
name: mcp-account-brief
description: Connect the Reverse Contact and CRM MCP servers to an AI client and have i
  t research an account and write a pre-call brief onto the CRM record. Use wh...
---

# Agent-written account briefs

## What you are building
## Prerequisites
## The tools, and how to call them   → Reverse Contact, HubSpot
## Build it                         → 4 steps
## Gotchas
## What breaks, from the vendors' own docs
## How to verify it works           → 5 checks
markdown Verified 2026-09-18
Drop it into Claude Code

mkdir -p .claude/skills/mcp-account-brief curl -o .claude/skills/mcp-account-brief/SKILL.md https://www.reversecontact.com/integrations-hub/skills/agent-account-briefs.md

For Codex, Cursor, or anything else that reads a file or fetches a URL, hand it https://www.reversecontact.com/integrations-hub/skills/agent-account-briefs.md directly. It is plain markdown with YAML frontmatter and no tool-specific syntax, so nothing in it depends on which agent is reading it.

3 of the joins in this chain have a full guide behind them, HubSpot and Reverse Contact and Slack and Reverse Contact among them. The chain crosses 3 of the seven layers a stack splits into, from the data layer outward, which is why no single vendor owns it.

What will bite

Documented by one of the vendors in the chain, and easier to design around than to discover.

An agent will fill a gap unless told not to
This is the whole risk and it is entirely addressable in the prompt. Tools that return null are doing the right thing; a model that treats a null as an invitation to estimate is not. Test it by asking for a brief on a company you know is thinly covered and checking what it says about headcount.
It writes to your CRM under your permissions
Which is convenient and means the blast radius is whatever you can do. Start it on notes rather than on properties. A note is additive and reviewable; a property write is a change to a field somebody else depends on.
Credits are consumed per call, invisibly
An agent exploring an account may make several enrichment calls where a pipeline would make one. The usage-check tool exists for exactly this, and it is worth asking for a figure after the first week rather than at the end of the month.
The live tools are asynchronous here too
The live variants return a task that the task-check tool polls. That works fine in an agent loop and it is slower than the cached variants. Use live only when freshness is the reason you are doing this at all.
It does not scale to a list
One account at a time is the shape. If the job is 5,000 accounts on a schedule, that is a pipeline and it belongs in the CRM data layer. Using an agent for that is slower, less predictable and more expensive.

All five are documented by a vendor rather than observed by us, which is the useful part: a limit you can read is a limit you can design around before it costs you a week.

What it costs to run

What each component bills on, and which line actually scales with your volume.

ComponentBillingWhat drives it
Reverse ContactPer creditEvery call the agent makes. Usually more per account than a pipeline would make, because an agent explores.
HubSpot connectorFreePublished by HubSpot. It runs under your own CRM permissions.
AI clientWhatever yours costsUnchanged by this workflow. No hosting, no scheduler, no instance.
BuildZeroTwo connections and a prompt. There is nothing to maintain when an API version moves.

The variable nobody expects is how many enrichment calls an agent makes while exploring one account. Check your usage after a week rather than guessing, and put a ceiling on it if the answer surprises you.

Built by

Reverse Contact

We sell B2B people and company data, and we publish this directory. Every tier requirement, limit and sync direction in this chain was read from a vendor's own documentation, through the integration records each step rests on. Nothing on it is restated from a press release or a third-party listing.

Where a vendor publishes no public figure, this page says so rather than estimating, which is why no page here prints a cost per enriched record. Reverse Contact appears in this directory as a data provider and publishes it. No other vendor pays for placement, and every link out of a listing or a citation carries nofollow, including the ones that point at Reverse Contact. The only followed links on this page are the two calls to action and the reviewer.

Tier requirements, sync behaviour and documented limits change without notice. Confirm against your own contract before you budget. What has changed since · How we checked, and what we will not claim

This workflow runs on a data layer you can test free

Reverse Contact resolves an email, a domain or a profile URL into a verified people and company record, with a confidence score your logic can branch on. The free trial takes no card and runs until the 500 credits a new account starts with are spent, which is enough to run this end to end before anyone signs anything.

Free trial, no card required, and 500 credits to start.

Questions

Which AI clients does this work in?

Reverse Contact lists Claude.ai, Claude Desktop, Cowork, Claude Code, ChatGPT, Codex, Cursor and Gemini CLI, plus any stdio-based MCP client through mcp-remote. The server speaks streamable HTTP with JSON-RPC 2.0. In practice, if your client supports remote MCP servers, it works.

Is this a replacement for enrichment in the CRM?

No, and treating it as one is the main way to be disappointed by it. Enrichment in the CRM is a scheduled loop over every record that needs a field. This is on-demand research over one account, producing prose rather than properties. Teams that need both run both, and the two barely interact.

How do I stop it writing something wrong to the CRM?

Two controls. In the prompt, tell it to preserve nulls and attribute each fact to its source. In the setup, start it on notes rather than properties, so everything it produces is additive and reviewable rather than a change to a field your routing depends on. Move to property writes only once you have read a month of notes and trust them.

Does this replace the integration entirely?

For this class of job, yes, and that is the interesting part. If an agent can call both APIs under your own credentials, the integration between them does not need to exist and nobody maintains it. What it does not replace is the scheduled, high-volume, two-way sync, which is still a pipeline and should stay one.

Tier requirements, sync behaviour and limits read from each vendor's own documentation in September, October and November 2026. Vendors change these without notice.

Reverse Contact appears in this directory as a data provider and publishes it. Every other tool is here on merit and nobody pays for placement.

Ask an assistant