ChatGPT and Claude Agents for GTM Engineers: The Complete Directory

Source: github.com/milisp/awesome-chatgpt-claude-agents

Filtered for: GTM engineers, SaaS founders, Directors, Owners, and Managers at B2B SaaS and agency companies who own outbound, lead gen, or GTM functions.

This directory covers the agents, subagents, frameworks, and tools from the curated list that are directly relevant to building agentic GTM systems. Everything here is open source, publicly available, and actively maintained.

What counts as relevant for GTM engineers:

• Agents that automate research, outreach, content, or pipeline work
• Coding agents that build GTM tools faster
• Multi-agent orchestration frameworks for running parallel workflows
• MCP-compatible tools for connecting Claude to live data
• Claude Code subagent collections with business and productivity agents

Why agents beat prompts for GTM

A prompt is a one-shot request. An agent is a workflow that runs, checks its own output, makes decisions, and completes a task end to end.

For GTM engineers, the difference matters:

Prompt: "Write a cold email for this prospect."

Result: one email, which you then paste into your CRM manually.

Agent: "Research [COMPANY], find the decision maker, pull their recent LinkedIn activity, write a cold email that references a specific signal, push the draft to Instantly, and log the contact to CRM."

Result: the full workflow runs. You review and approve.

The 2025 trend across the entire agent ecosystem: domain-specific specialization. Industry-focused agents that know your specific workflow, your specific tools, and your specific ICP beat general-purpose models every time.

The 2025 agent landscape for GTM

Four trends from the repo that matter for GTM engineers:

1. Model Context Protocol (MCP): standardizing how AI connects to external tools. Every CRM, email platform, LinkedIn tool, and data enrichment provider will have an MCP connector. When they do, your agent can use them all from one session.
2. Multi-modal agents: combining text, image, and data processing in one workflow. For GTM: research + writing + graphic generation in a single agent run.
3. Autonomous coding: agents that build and fix their own tools. For GTM engineers: if a skill does not exist yet, you can describe it and an agent builds it.
4. Domain-specific specialization: industry-focused agents that know your workflow beat general-purpose models. The best agents in this directory are purpose-built for specific GTM functions, not general use.

ChatGPT agents relevant to GTM engineers: research, productivity, content, and business automation.

Filtered from the full list. Only includes agents with direct GTM application.

GPT Researcher

Repo: github.com/assafelovic/gpt-researcher

Stars: 14,000+

What it is: an autonomous GPT-based agent that does comprehensive online research and produces structured research reports.

GTM use cases:

• Account research before outreach: "Research [COMPANY] and produce a structured brief covering what they do, their tech stack signals, recent news, decision makers, and why now is the right time to reach out."
• Competitor intelligence: "Research [COMPETITOR] and produce a competitive brief covering their positioning, pricing signals, G2 review themes, and what customers complain about."
• Market research for ICP refinement: "Research [INDUSTRY] in 2026 and identify the top 5 signals that indicate a company is in buying mode for [CATEGORY]."
• Signal-based outreach research: "Find recent funding, hiring, or news signals for companies matching [ICP DESCRIPTION] from the last 30 days."

Why it beats a basic prompt: it runs multiple searches, synthesizes across sources, and produces a structured output you can act on directly. A basic prompt produces one response from one source.

TaskWeaver

Repo: github.com/microsoft/TaskWeaver

Stars: 5,000+

What it is: a code-first agent framework for seamlessly planning and executing data analytics tasks. Converts user requests into executable Python code, runs it, and returns structured results.

GTM use cases:

• Pipeline analytics: connect your CRM export and ask "which stage has the longest average time and why?" It writes and runs the analysis.
• Email campaign performance: "Analyse my Instantly export and tell me which subject line patterns produce the best positive reply rates."
• Lead list scoring: "Score this CSV of 500 prospects against my ICP criteria and rank them."
• Attribution analysis: "Analyse this HubSpot export and tell me which first-touch channels produce the highest close rates."

Why it is relevant: most GTM analytics involves data that lives in CSV exports. TaskWeaver turns natural language questions into running code against those files.

ChatGPT-Code-Review (for GTM tool auditing)

Repo: github.com/anc95/ChatGPT-CodeReview

What it is: a code review bot powered by ChatGPT. Reviews PRs and code changes automatically.

GTM use cases:

• Auditing Claude Code skills before deploying them to your team: "Review this SKILL.md and flag any logic errors, missing edge cases, or places where the workflow might fail."
• Reviewing n8n workflow JSON before importing: "Review this n8n workflow and flag any steps that could fail silently or send messages without an approval gate."
• Auditing API integration code for CRM or outreach tools.

For GTM engineers building their own agent stack: code review agents catch problems before they run against live data or send live messages.

AI-Commit (for GTM code repos)

Repo: github.com/guanguans/ai-commit

What it is: automatically generates conventional commit messages with AI. Reads your git diff and writes a clean commit message.

GTM use case: if you are maintaining a skills library, an n8n workflow repo, or a Clay template repo with your team, this agent keeps your commit history clean and searchable without requiring anyone to write commit messages manually.

Claude Code subagents relevant to GTM engineers. These run inside Claude Code as specialized sub-agents that Claude can spin up to handle specific parts of a workflow.

All collections are open source. Install commands included.

VoltAgent: 100+ production-ready subagents

Repo: github.com/VoltAgent/awesome-claude-code-subagents

What it is: a production-ready collection of 100+ specialized AI subagents for full-stack development, DevOps, data science, and business operations. These are the most comprehensive single collection in the directory.

GTM-relevant subagents in this collection:

• Business operations agents: planning, reporting, analysis
• Data analysis agents: CSV processing, pipeline analysis, cohort analysis
• API integration agents: connect to external tools and data sources
• Code review agents: audit skills and workflows before deploying
• Documentation agents: write and maintain GTM playbooks and SOPs

# Install
git clone https://github.com/VoltAgent/awesome-claude-code-subagents.git
cp -r awesome-claude-code-subagents/agents/* ~/.claude/agents/

How subagents work: in Claude Code, when you run a complex task, Claude can spin up specialized sub-agents to handle parts of it in parallel. A GTM workflow might spin up a research agent, a writing agent, and a CRM update agent simultaneously.

wshobson/agents: 75 specialized subagents

Repo: github.com/wshobson/agents

What it is: 75 specialized subagents that extend Claude Code's capabilities for specific domains. More focused than the VoltAgent collection, with higher per-agent quality.

GTM-relevant subagents:

• Market research agents
• Business analysis agents
• Writing and content agents
• Data processing agents

# Install
git clone https://github.com/wshobson/agents.git
cp -r agents/agents/* ~/.claude/agents/

0xfurai: 100+ development subagents

Repo: github.com/0xfurai/claude-code-subagents

What it is: a comprehensive collection of 100+ production-ready development subagents. Strong on technical workflows: debugging, testing, API integration, code generation.

GTM-relevant for engineers building their own stack:

• API integration agents for connecting to outreach tools, CRMs, and data enrichment providers
• Debugging agents for diagnosing why a skill or workflow failed
• Testing agents for validating that an outreach workflow behaves correctly before going live

# Install
git clone https://github.com/0xfurai/claude-code-subagents.git
cp -r claude-code-subagents/agents/* ~/.claude/agents/

toprank: SEO and Google Ads skills

Repo: github.com/nowork-studio/toprank

What it is: open-source Claude Code plugin with 9 SEO and Google Ads skills. Connects Google Search Console, PageSpeed Insights, and Google Ads API directly to Claude Code.

GTM use cases:

• If your GTM motion includes paid ads, this plugin lets Claude Code query your Google Ads performance, audit campaign structure, and suggest optimisations directly from the terminal.
• SEO skills for content teams running inbound alongside outbound.
• PageSpeed audits for landing pages used in outbound campaigns.

claude plugin marketplace add nowork-studio/toprank
claude plugin install toprank@nowork-studio

vijaythecoder: 24 agents for production features

Repo: github.com/vijaythecoder/awesome-claude-agents

What it is: 24 specialized agents working together to build production-ready features. Designed for coordinated multi-agent workflows where agents hand off between each other.

GTM application: useful for GTM engineers building features into their own products (SaaS founders) rather than using existing tools. If you are building outreach automation into your product, these agents accelerate development.

OpenPaw: 38-skill personal assistant

Repo: github.com/daxaur/openpaw

Install: npx pawmode

What it is: open-source CLI tool that turns Claude Code into a personal assistant with 38 skills. No daemon, no cloud, MIT licensed. Install with one command.

GTM use cases: the 38 skills cover productivity, research, writing, and scheduling use cases. For GTM engineers who want a lightweight assistant layer on top of Claude Code without setting up a full skills library.

npx pawmode

Multi-agent orchestration frameworks let you run multiple agents in parallel, hand off between them, and coordinate complex workflows. For GTM: research + enrichment + writing + CRM update running simultaneously.

CrewAI

Repo: github.com/joaomdmoura/crewAI

Stars: 20,000+

What it is: a framework for orchestrating role-playing, autonomous AI agents. Each agent has a role, a goal, and a set of tools. You define the crew, the tasks, and the process. CrewAI handles the coordination.

GTM use case: build a GTM research crew:

"Create a CrewAI workflow with 3 agents:

Agent 1 (Researcher): Research [COMPANY NAME] and produce a structured brief
covering decision makers, recent signals, tech stack, and buying intent indicators.

Agent 2 (Writer): Take the research brief and write a personalised first-touch email
for [TARGET TITLE] referencing the strongest signal found.

Agent 3 (Reviewer): Review the email against these rules: under 150 words, no em dashes,
no generic openers, CTA is a question. Return pass/fail with rewrites for any failures.

Run this crew for: [LIST OF COMPANY NAMES]"

Why CrewAI for GTM: it handles the coordination between agents so you do not have to chain prompts manually. Each agent in the crew specialises in one part of the workflow.

Autogen

Repo: github.com/microsoft/autogen

Stars: 32,000+

What it is: Microsoft's framework for enabling multi-agent conversational AI applications. Agents talk to each other to solve problems collaboratively.

GTM use case: build an Autogen group for pipeline management:

"Set up an Autogen group with:
- Pipeline Manager: reviews the CRM pipeline weekly and flags at-risk deals
- Research Agent: when a deal is flagged, researches the account for new signals
- Strategy Agent: given the research, recommends a specific next action
- Writer Agent: drafts the next outreach message referencing the signal

Run the group against this week's pipeline data: [PASTE OR PULL FROM CRM]"

MetaGPT

Repo: github.com/geekan/MetaGPT

Stars: 44,000+

What it is: a multi-agent framework that turns a one-line requirement into PRD, design, tasks, and code. Simulates a software company with product manager, engineer, and QA roles.

GTM use case for SaaS founders and technical GTM engineers: MetaGPT is the fastest way to scaffold a GTM tool from scratch. Give it "build a cold email personalisation tool that takes a LinkedIn URL and outputs a first-touch email" and it produces the full spec and code.

Also useful for: building internal GTM tools (CRM dashboards, sequence builders, signal monitors) without a full engineering team.

Swarm (OpenAI)

Repo: github.com/openai/swarm

What it is: an educational framework for lightweight multi-agent orchestration by OpenAI. Simpler than CrewAI or Autogen. Good starting point for GTM engineers who want to understand multi-agent coordination before building more complex systems.

GTM use case: use Swarm to prototype a simple outbound workflow before building it out in Claude Code or CrewAI:

# Simple Swarm example: research → write → review
Agent 1: Research the prospect and return key signals
Agent 2: Write a cold email using the signals
Agent 3: QA the email and return pass/fail
# Handoffs: 1 → 2 → 3 → you

Development platforms relevant to GTM engineers building their own tools or running heavy agentic workloads.

OpenDevin (now OpenHands)

Repo: github.com/OpenDevin/OpenDevin

Stars: 61,000+

What it is: open-source initiative replicating autonomous AI software engineer Devin. Can write code, run it, debug it, and iterate without human intervention at each step.

GTM use cases for technical GTM engineers and SaaS founders:

• "Build me a Clay-style enrichment workflow that takes a CSV of company names and returns decision maker emails using Apollo and Hunter in a waterfall."
• "Build me a LinkedIn signal monitor that watches a list of target accounts and notifies me via Slack when any of them post about [TOPIC]."
• "Build me a dashboard that pulls my Instantly campaign stats and shows open rate, reply rate, and positive reply rate by subject line variant."

For SaaS founders: this is how you build GTM tools with a team of one.

E2B

Repo: github.com/e2b-dev/e2b

What it is: an open platform for running AI agents in sandboxed cloud environments. AI agents can write and execute code in isolated environments without security risk.

GTM use cases:

• Run data analysis agents against your pipeline data in a secure sandbox
• Execute enrichment scripts against lead lists without running code locally
• Test outbound automation workflows in isolation before deploying to live accounts

For GTM engineers who need to run agent-generated code against real data without the risk of running arbitrary code on their machine.

GPT-Engineer

Repo: github.com/AntonOsika/gpt-engineer

Stars: 52,000+

What it is: specify what you want it to build, the AI asks for clarification, and then builds it. One of the original code generation agents.

GTM use cases:

• Build a custom cold email personalisation script in Python
• Build a signal monitor for a specific data source
• Build a lightweight CRM data cleaning tool
• Build an Airtable or Notion automation that does not have a native integration

For GTM engineers who are not daily coders but need custom tools: describe what you want in plain language, GPT-Engineer asks the right questions, and builds it.

Lobe Chat

Repo: github.com/lobehub/lobe-chat

Stars: 63,000+

What it is: an open-source UI framework for building ChatGPT and LLM-based chat applications. Self-hostable. Supports multiple models.

GTM use case: GTM teams who want a self-hosted chat interface for their team. Build a custom GPT/Claude interface with your GTM skills and CLAUDE.md baked in, hosted internally. Every team member gets the same context-aware AI without paying per-seat for Claude Team.

The most relevant guides, tutorials, and research from the directory for GTM engineers building agentic systems.

OpenAI: Practical Guide to Building Agents

URL: cdn.openai.com/business-guides-and-resources/a-practical-guide-to-building-agents.pdf

What it covers: step-by-step guide on tools, guardrails, and agent architecture. Written for practitioners, not researchers.

Key concepts relevant to GTM engineers:

• Orchestration: how to coordinate multiple agents without them conflicting
• Guardrails: how to prevent agents from sending messages or making changes without approval
• Tool use: how agents call external APIs reliably
• Memory: how agents retain context across sessions

Most useful section for GTM: the guardrails section. Every GTM agent should have an approval gate before anything touches LinkedIn, email, or CRM. This guide covers how to build that correctly.

Building AI Agents with Claude (Medium)

URL: medium.com/aingineer/building-ai-agents-with-claude-3-7-a-comprehensive-guide-part-1-07d9df717a04

What it covers: comprehensive guide on agent components for Claude: memory, tool integration, planning, and execution loops.

GTM application: understanding how Claude's memory works helps you structure CLAUDE.md correctly and know what to include in your session context vs what to load per-request.

Turn Claude into Powerful AI Agents (YouTube)

URL: youtube.com/watch?v=p0pR_zq-85M

What it covers: how to automate workflows using MCP and Claude. One of the clearest practical explanations of how MCP connects Claude to external tools for real workflow automation.

GTM application: the same principles shown in this video apply to connecting Claude to Instantly, Smartlead, HubSpot, Salesforce, and LinkedIn via their MCP servers or API.

The Ultimate Guide to AI Agents for PMs

URL: productcompass.pm/p/ai-agents

What it covers: frameworks, tools, and resources for building agents. Written for product managers but directly relevant to GTM engineers who approach agent building from a workflow design perspective rather than a coding perspective.

GTM application: the frameworks in this guide help you design your GTM agent system before you build it. What are the stages? What are the handoffs? Where do humans need to stay in the loop?

Everything from this directory combined into a GTM-specific agent stack you can build today.

The GTM agent stack (what to actually install)

For most GTM engineers, this is the practical stack:

Core: Claude Code with Sonnet 5

# Model string
claude-sonnet-5

Subagent collections (pick one to start):

# Option A: VoltAgent (100+ agents, broadest coverage)
git clone https://github.com/VoltAgent/awesome-claude-code-subagents.git
cp -r awesome-claude-code-subagents/agents/* ~/.claude/agents/

# Option B: wshobson (75 agents, higher per-agent quality)
git clone https://github.com/wshobson/agents.git
cp -r agents/agents/* ~/.claude/agents/

# Option C: OpenPaw (38 skills, one command, no setup)
npx pawmode

Research agent (for account research and signal monitoring):

# GPT Researcher (run via Python)
git clone https://github.com/assafelovic/gpt-researcher.git
cd gpt-researcher && pip install -r requirements.txt --break-system-packages

Orchestration (for multi-agent GTM workflows):

# CrewAI (for parallel research + writing + review workflows)
pip install crewai --break-system-packages

# Or stay inside Claude Code and use the plan mode + sub-agent approach
# Claude Code handles orchestration natively via Workflows

The 5 GTM agent workflows to build first

1. Account research agent

"I need a full account brief for [COMPANY NAME] before I reach out.
Run GPT Researcher on them and produce:
- What they do in one sentence
- Key decision makers and their titles
- Recent signals from the last 90 days
- Tech stack signals from job postings
- Recommended angle for first touch"

1. Signal-led outreach agent

"Build a CrewAI crew:
Agent 1 (Researcher): Given [COMPANY], find 2 verified signals from last 90 days
Agent 2 (Writer): Write a first-touch email referencing the strongest signal. Under 150 words.
Agent 3 (Reviewer): Score the email 0-100. Flag anything under 15 per dimension.
Return the email only if it scores 80+. Otherwise return the reviewer's rewrite."

1. Pipeline analysis agent

"Use TaskWeaver to analyse my pipeline data: [PASTE CSV OR PULL FROM CRM]
Question: which stage has the highest deal drop-off and what is the average deal age at that stage?
Return a structured analysis with the top 3 deals to prioritise this week."

1. Competitor intelligence agent

"Run GPT Researcher on [COMPETITOR]. Produce:
- Their positioning and messaging (what they claim to be best at)
- Pricing signals (any public pricing or tier information)
- G2 review themes: what customers love and what they complain about
- Recent news or product announcements
- 3 talking points I can use in competitive conversations"

1. GTM tool builder

"Use GPT-Engineer or OpenDevin to build:
A Python script that takes a CSV of LinkedIn URLs, looks up each profile via the Linked API CLI,
and returns a CSV with name, title, company, company size, and email.
Include: 30-day dedup check, rate limiting (max 50 lookups per run), and error handling."

What to skip from the directory

Not relevant for GTM engineers:

• AutoGPT (177k stars but too general, too unpredictable for GTM workflows that send live messages)
• ChatGPT-Academic (research paper tool, not GTM)
• Paper-QA (academic, not GTM)
• Security-GPT / CyberGPT / PentestGPT (cybersecurity, not GTM)
• FinGPT / TradingGPT (finance trading, not GTM)
• CAMEL / ChatDev (research frameworks, too experimental for production GTM work)
• Storyteller (creative writing, not GTM)

Rule of thumb: if an agent is built for general autonomy (AutoGPT) or for a domain that is not business/sales/research (security, trading, academic), skip it. GTM agents need to be reliable, specific, and have approval gates. General-purpose autonomous agents are not safe to use in a workflow that sends live messages or updates live CRM data.