AI GTM Agent

An agent-agnostic go-to-market operating layer for founders, agencies, and revenue teams — connecting market research, ICP definition, positioning, launch planning, sales motion, channel execution, and performance review.

Part of Agency Agents OS, an open ecosystem of specialized AI agents for real business work.

What this covers

This repo is the cross-functional GTM umbrella for work that spans several specialized capabilities. It keeps one shared view of the product, audience, market, offer, funnel, channels, goals, evidence, and constraints, then routes the right parts of the work to sibling umbrellas:

• ai-research-agent for market and audience evidence.
• ai-seo-agent and ai-aeo-geo-agent for organic and AI-search visibility.
• ai-marketing-agent for email, content campaigns, and product marketing.
• ai-sales-agent for prospecting, enrichment, verification, and outreach drafts.
• ai-ads-agent for paid-media intelligence, testing, and ROAS analysis.
• ai-social-agent for social publishing, listening, creator discovery, and trend research.
• ai-analytics-agent for reports and dashboard specifications.
• The media, e-commerce, reputation, and competitor-intelligence umbrellas for the supporting assets and evidence those workflows need.

It is deliberately not another generic sales prompt pack, CRM, sequencer, ad budget manager, or autonomous sending service. The GTM layer coordinates the specialists and turns their outputs into one decision-ready plan.

Sub-agents

Agent	What it does	Status
GTM Project Setup	Establishes reusable product, market, ICP, funnel, channel, and goal context	Coming Soon
GTM Strategist	Routes a broad GTM question to the smallest useful set of sibling workflows and synthesizes the result	Coming Soon
GTM Launch Orchestrator	Turns an approved strategy into a cross-channel launch plan, dependency map, calendar, and approval gates	Coming Soon
GTM Performance Review	Joins available funnel and channel evidence to identify bottlenecks and propose the next experiments	Coming Soon

How the workflows fit together

GTM Project Setup
        |
        v
GTM Strategist -----> Research / SEO / Marketing / Sales / Ads / Social / Analytics
        |
        v
GTM Launch Orchestrator -----> sibling asset and execution workflows
        |
        v
GTM Performance Review -----> findings, bottlenecks, and next experiments

The dependency order is intentional. Establish the context before collecting broad evidence, decide the strategy before building a launch calendar, and review performance only after the reporting window and metric definitions are known.

Shared GTM project context

When a team has an ongoing GTM project, the host agent may keep the confirmed context and local reports under:

.gtm/
  project.md
  reports/YYYY-MM-DD/<run-slug>.md
  sources/YYYY-MM-DD/<run-slug>/<source-name>.json

The project context should contain only confirmed or explicitly user-supplied information:

• product, offer, business model, and launch stage
• target markets, language, geography, and sales motion
• ICP, segments, buyer roles, pain points, and exclusions
• positioning, proof points, objections, and known alternatives
• funnel stages, goals, success metrics, and conversion definitions
• channels, budget constraints, launch dates, owners, and approval rules
• confirmed competitors and links to supporting source records
• known data gaps and the date the context was last reviewed

Required capabilities

This repo does not assume a new monolithic gtm.* API. It consumes existing and planned capability contracts from the sibling umbrellas, including:

• research.web_search and research.audience_profile
• company.enrich, people.search, email.verify, and outreach.draft_message
• SEO, AEO/GEO, YouTube, local, and competitor-intelligence capabilities
• marketing.campaign_analytics and analytics.ga4_report
• ads.ad_library_search, ads.creative_metrics, and ads.campaign_performance
• social.read_posts, social.analytics, and social.list_accounts
• reputation.news_search, reputation.review_search, and social.sentiment_analysis

The host agent should use the live provider schema as the final source of truth. A capability that is unavailable, partial, or supplied through a user file must be labeled that way in the report; it must not be replaced with an estimate.

Setup

1. Create a Muapi account and API key at muapi.ai when live external research is required.
2. Load gtm-project-setup/SKILL.md for a new or stale GTM project.
3. Load gtm-strategist/SKILL.md for a broad GTM question.
4. Load gtm-launch-orchestrator/SKILL.md when the strategy is approved and needs to become an execution plan.
5. Load gtm-performance-review/SKILL.md after a reporting window has closed and comparable metrics are available.

Every skill is plain Markdown and can be loaded by Claude, Codex, Cursor, Windsurf, an MCP client, or another runtime that supports project instructions. The host agent remains responsible for reasoning, tool calls, approvals, and local report persistence.

Read-only and approval boundaries

All four workflows are read-only or draft-only. They may prepare plans, reports, asset briefs, and handoffs, but they never:

• send email or LinkedIn messages
• publish or schedule social content
• launch or replace ads
• change budgets, pricing, CRM records, or external websites
• contact prospects, customers, creators, reviewers, or journalists

Any downstream action belongs to the relevant sibling workflow and requires its stated human approval boundary.

Status and limitations

All four sub-agents are Coming Soon. The orchestration design is ready, but several of the cross-functional data capabilities it depends on — research, B2B enrichment, marketing analytics, ad performance, social analytics, and reputation data — are not yet live across the sibling umbrellas. Until those capabilities are available, the host may work from user-supplied briefs or exports, but must label those inputs clearly and must not present them as live Muapi data.

Contributing

See Agency Agents OS CONTRIBUTING.md.

License

MIT