# Swarms: Build a 3-Agent Research Team That Writes a Client Brief > Build a Swarms SequentialWorkflow in Python: a Researcher, an Analyst and a Writer turn discovery-call notes into a one-page client brief. Free lesson. Source: https://learn.ynteractive.com/content/swarms-multi-agent-client-brief · Updated 2026-11-06 · Free from Stepthrough (https://learn.ynteractive.com) ## Practice it step by step [Start the free practice](https://learn.ynteractive.com/modules/swarms-client-brief-team) [Take the free module: why it matters, practice and a quick check →](https://learn.ynteractive.com/courses/ai-agents-101/12) You do every step yourself in a practice copy of Swarms. It checks your work as you go, and “Do it for me” finishes any step you get stuck on. Free, about 9 minutes, and nothing touches your real accounts. [Open full screen](https://learn.ynteractive.com/modules/swarms-client-brief-team) Free · Live now · [AI Automation for Consulting & Agencies](https://learn.ynteractive.com/content/industry/consulting-and-agencies) · Swarms · Python · ~9 min · 12 steps · Advanced Updated November 6, 2026. Built against the tool’s current screens and checked step by step. [How we check this lesson](https://learn.ynteractive.com/content/about#how-we-check) Consultants and account managers lose an hour after every discovery call turning rough notes into a brief the team can act on, and the risks and open questions are the first things cut when time is short. Swarms is a free, open-source [Python](https://learn.ynteractive.com/content/glossary/python) framework for building teams of [AI agents](https://learn.ynteractive.com/content/n8n-ai-agent-tutorial-google-sheets). In this free starter lesson you install it, add an [OpenAI](https://learn.ynteractive.com/content/glossary/openai) key, write three agents with one job each (a Researcher, an Analyst and a Writer), chain them in a SequentialWorkflow, and run them on a set of call notes to save a one-page client brief. The code follows the Swarms docs' Quickstart and is about 30 lines; you paste each part and the lesson explains it. ## At a glance | Question | Answer | | --- | --- | | What you build | Turn rough discovery-call notes into a one-page client brief with three AI agents: a Researcher pulls the facts, an Analyst finds the risks and open questions, a Writer writes the brief. About 30 lines of Python, straight from the Swarms docs. | | Tools | Swarms | | Time | About 9 minutes, 12 steps in 3 parts | | Level | Advanced | | Cost | Free | | Ways to learn it | [Practice it step by step](https://learn.ynteractive.com/modules/swarms-client-brief-team) | ### The 3 parts, in short 1. Set up Swarms: Open the call notes, Install Swarms, Open .env, Add your OpenAI key. 2. Three agents, one job each: Open client_brief.py, Agent 1: the Researcher, Agent 2: the Analyst, Agent 3: the Writer, and 1 more step. 3. Run the team: Feed it the notes, Run it, Open the brief. **Start here first:** [Granola: Client Call Notes Written for You, With No Meeting Bot](https://learn.ynteractive.com/content/granola-client-call-notes-ai-notepad) (module 1 of AI Automation for Consulting & Agencies) **Part of:** [AI Agents 101](https://learn.ynteractive.com/content/paths/ai-agents-101) (step 16 of 16) **Part of:** [AI Apps Worth Knowing](https://learn.ynteractive.com/content/paths/partner-tools-ai-apps) (step 132 of 140). Next up: [TryPost: Turn One Blog Post into a Week of Scheduled Social Posts](https://learn.ynteractive.com/content/trypost-n8n-blog-post-to-week-of-social-posts) ## The business case: an hour per discovery call Picture Sam, a consultant at Northgate Advisory, a made-up firm. After a discovery call with a freight broker, Sam has a page of rough notes: who was on the call, the goal, a budget, a deadline, a worry from the sales lead and where the rate data lives. Turning that into a brief the team can use takes about an hour. With example numbers you should replace with your own, four discovery calls a week is half a working day. Splitting the job between three agents keeps each step simple. The Researcher only pulls facts. The Analyst only reads those facts and names priorities, risks and open questions. The Writer only writes the brief with the same headings every time. You read the result against your notes and fix anything wrong in five minutes. ## Who this is for, and what you need Consultants, account managers and operations leads who are comfortable copying a few lines of Python. To run it for real you need Python 3.10 or newer, the Swarms package (pip3 install -U swarms) and an API key for a model provider such as OpenAI. The lesson itself runs in a practice editor in your browser, so you need nothing installed to try it. ## How a SequentialWorkflow works A SequentialWorkflow runs agents in the order you list them. The first agent gets your task, here the call notes, and each later agent gets the conversation so far, so the Analyst works from the Researcher's facts and the Writer works from both. The default output_type is "dict", a list of every message; "final" returns only the last agent's answer, which is what you want to save as the brief. If one part of the brief is weak, you change one agent's system_prompt and leave the rest alone. Swarms also has a ConcurrentWorkflow for agents that should work at the same time, and options such as drift detection that score the final answer against the task and rerun if it drifts. ## Keep it safe - Keep the API key in .env, never in the script, and never commit .env to Git. - Call notes can hold client details. Check your firm's rules before you send them to a model provider. - The brief is a draft for you. Agents can misread a note or fill a gap with a guess, so read it against your notes before you share it. ## What this saves you - Most of the hour after each discovery call. - A brief with the same headings every time, so risks and open questions are never skipped. - What it does not do: it does not research the client on the web (these agents have no tools), and it does not send anything to the client. ## How this lesson was built This is a starter lesson built from the Swarms public docs (Installation, Environment Setup, Quickstart and Sequential Workflow), read on 6 November 2026. The code uses the docs' own pattern and parameter names. We did not run it against a real model for this lesson, so the terminal output and the brief shown are examples, not captured results. A full Swarms course, tested in a real run, is planned. ## Links - [Swarms docs: Quickstart](https://docs.swarms.world/quickstart): the Agent and SequentialWorkflow example this lesson's code is based on. - [Swarms docs: Sequential Workflow](https://docs.swarms.world/architectures/sequential-workflow): every SequentialWorkflow option, including output_type, team_awareness and drift detection. - [Swarms docs: Environment Setup](https://docs.swarms.world/environment-setup): where API keys go (.env) and the other settings Swarms reads. - [Swarms on GitHub](https://github.com/kyegomez/swarms): the open-source code. ## Try it yourself, free Reading the steps is a start. Doing them is how it sticks. The interactive module walks you through every click in a practice copy of Swarms, checks each step, and never touches your real accounts. [Start “Swarms: Build a 3-Agent Research Team That Writes a Client Brief”](https://learn.ynteractive.com/modules/swarms-client-brief-team) ## Step by step: Swarms: Build a 3-Agent Research Team That Writes a Client Brief All 12 steps of the interactive module, in order. In the module you watch each one, then do it yourself in a practice copy that checks your work. ### 1. Set up Swarms 1. **Open the call notes.** You are a consultant at **Northgate Advisory**, a made-up firm. After every discovery call you turn your notes into a one-page client brief for the team, and it takes an hour. This project folder has today's notes. Click **call_notes.txt**. You are a consultant at Northgate Advisory. After every discovery call, you turn your notes into a one-page client brief, and it takes an hour. 2. **Install Swarms.** Messy, like real notes. **Swarms** is a free, open-source Python framework for teams of AI agents. Install it in the terminal with the command from its docs: **pip3 install -U swarms**. Paste in the command, then press **Enter**. Messy, like real notes. Swarms is a free, open-source Python framework for teams of AI agents. Install it with the command from its docs. 3. **Open .env.** Each agent calls an AI model, so Swarms needs your model provider's API key. The docs keep it in a **.env** file in the project folder. Click **.env**. Each agent calls an AI model, so Swarms needs your API key. The docs keep it in a dot env file. 4. **Add your OpenAI key.** Add your OpenAI key on one line. Here it is the docs' placeholder: Paste in the key. Never paste a real key into a script or share the .env file. Add your OpenAI key. Here it's the placeholder from the docs. Never put a real key in the script itself. ### 2. Three agents, one job each 5. **Open client_brief.py.** Now the team. Click **client_brief.py**, an empty file. Now the team. It's empty. 6. **Agent 1: the Researcher.** An **Agent** is one AI worker with one job, set in its **system_prompt**. The Researcher pulls the facts out of the notes and nothing else. Paste in the code. An agent is one AI worker with one job, set in its system prompt. The Researcher pulls the facts out of the notes, and nothing else. ```text from swarms import Agent, SequentialWorkflow researcher = Agent( agent_name="Researcher", system_prompt="Read the call notes. List every fact about the client: who they are, what they do, their goal, budget, deadline and who decides. Facts only, no advice.", model_name="gpt-5.4", max_loops=1, ) ``` 7. **Agent 2: the Analyst.** The Analyst never sees your notes. It reads the Researcher's facts and finds the priorities, the risks and the questions you still need to ask. Paste in the code. The Analyst reads the Researcher's facts, and finds the priorities, the risks, and the questions you still need to ask. ```text analyst = Agent( agent_name="Analyst", system_prompt="From the facts, name the client's top 3 priorities, the biggest risks to the project, and the questions we still need to ask.", model_name="gpt-5.4", max_loops=1, ) ``` 8. **Agent 3: the Writer.** The Writer turns all of that into the brief, with the same headings every time, so your team always knows where to look. Paste in the code. The Writer turns it into the brief, with the same headings every time. ```text writer = Agent( agent_name="Writer", system_prompt="Write a one-page client brief in Markdown with the headings Client, Goal, Budget and deadline, Priorities, Risks, Open questions and Next steps. Plain English, short bullets.", model_name="gpt-5.4", max_loops=1, ) ``` 9. **Line them up.** A **SequentialWorkflow** runs the agents in order: each one's answer becomes the next one's input. **output_type="final"** returns only the Writer's answer. Paste in the code. A Sequential Workflow runs the agents in order. Each answer becomes the next agent's input. Output type final returns only the Writer's answer. ### 3. Run the team 10. **Feed it the notes.** Last part: read **call_notes.txt**, run the team on it, and save the result as **client_brief.md**. Paste in the code. Last part. Read the call notes, run the team, and save the result as client brief dot m d. 11. **Run it.** In the terminal, run **python client_brief.py**. Paste in the command, then press **Enter**. Each agent answers in turn: Researcher, then Analyst, then Writer. Run the script. Each agent answers in turn. 12. **Open the brief.** A new file appeared. Click **client_brief.md**. A new file appeared. ## Frequently asked questions **What is Swarms?** Swarms is a free, open-source Python framework for building AI agents and teams of agents. You define each Agent with a name, a system prompt and a model, then combine them in workflows such as SequentialWorkflow or ConcurrentWorkflow. **How do I install Swarms?** Run pip3 install -U swarms with Python 3.10 or newer. The docs also show uv, poetry and installing from source. Put your model provider's API key, for example OPENAI_API_KEY, in a .env file in your project folder. **What does output_type="final" do in a SequentialWorkflow?** It returns only the last agent's answer as text. The default, "dict", returns every message in the run as a list of role and content entries. **Do I need to know Python for this lesson?** No. You paste each part of the code and the lesson explains it. To run it on your own computer you need Python installed and an API key. **Is this lesson free, and was it tested in a real run?** The lesson is free and runs in a practice editor with no account. It is a starter lesson built from the Swarms docs; the code was not run against a real model for it, and the output shown is an example. [Start this module free](https://learn.ynteractive.com/modules/swarms-client-brief-team) - Previous in AI Automation for Consulting & Agencies: [End Every Email Claude Drafts With Your Saved Signature using Claude and Signatoro](https://learn.ynteractive.com/content/claude-signatoro-connector-signature) - Next in AI Automation for Consulting & Agencies: [Add Your Company Email Signature to Every Zoho Mail Message with Signatoro](https://learn.ynteractive.com/content/signatoro-zoho-mail-signature) ## More in AI Automation for Consulting & Agencies - [Granola: Client Call Notes Written for You, With No Meeting Bot](https://learn.ynteractive.com/content/granola-client-call-notes-ai-notepad): Take a client call with Granola: jot a few lines, let it write the notes from the transcript, check the budget against its source, draft the follow-up email and share… - [Beautiful.ai: A Client Review Deck from Your Brief in Under an Hour](https://learn.ynteractive.com/content/beautiful-ai-client-review-deck-create-with-ai): Build a client quarterly review deck with Beautiful.ai Create with AI: brief it with the real numbers, fix the outline before generating, check every figure and share a… - [Turn a Client Kickoff Call into a Proposal Outline with ChatGPT (Agencies & Consultants)](https://learn.ynteractive.com/content/chatgpt-kickoff-call-proposal-outline-agency): Turn the transcript of a client kickoff call into the scope, budget and timeline the client stated, an action items table, open questions, and a recap email with a… - [Handle Scope Creep Without Losing the Client using ChatGPT](https://learn.ynteractive.com/content/chatgpt-scope-creep-email-reply-agency): A client asks for "one small extra" and a new language, and wants the launch date to hold. Give ChatGPT the thread and the project facts, catch the free work and the… - [Research a Prospect's Market with Claude Research Before a Pitch (Consultants & Agencies)](https://learn.ynteractive.com/content/claude-research-prospect-competitor-scan-agency): Turn on Research before a pitch, have Claude check what a prospect's competitors offer online, read the cited report, check a source and turn it into talking points for… - [Prompt Chaining for a Client Campaign Plan: Build It in Steps with ChatGPT (Agencies)](https://learn.ynteractive.com/content/prompt-chaining-client-campaign-plan-agency): Ask ChatGPT for a client's whole 4-week campaign in one prompt, then build a better one in four steps: audiences, offers, a calendar table checked against the client's… [All 19 modules in AI Automation for Consulting & Agencies](https://learn.ynteractive.com/content/industry/consulting-and-agencies) ## Same business, other tools Northgate Advisory shows up in these modules too, built with different tools. - [OLVA: A Private AI Assistant on Your Client Calls](https://learn.ynteractive.com/content/olva-ai-meeting-assistant-client-calls): Give OLVA your client notes, get private answers and fact checks during a Zoom call, then turn the meeting into minutes, a follow-up email and a PowerPoint deck. ## Terms in this lesson Plain-English definitions, each with an example from a free lesson. Discovery call · [Python](https://learn.ynteractive.com/content/glossary/python) · [AI agent](https://learn.ynteractive.com/content/glossary/ai-agent) · [OpenAI](https://learn.ynteractive.com/content/glossary/openai) · Sales lead · [API](https://learn.ynteractive.com/content/glossary/api) [The full AI glossary](https://learn.ynteractive.com/content/glossary) ## Browse related tutorials [AI for Consulting & Agencies (19)](https://learn.ynteractive.com/content/industry/consulting-and-agencies)