Agent-Native MarTech

What It Is

Agent-native MarTech inverts the traditional tool model: instead of a dashboard a human logs into, the vendor exposes marketing channels (Search Console, Google Ads, GA4, ESPs, Discord, …) as MCP servers or APIs that the customer's own agent drives, and packages expertise as skills — codified playbooks like /seo-audit or /deliverability-check — rather than UI features. The reference example is Cogny Solo: $9/month for unmetered MCP access to 13+ channels, one-command install into Claude Code/Codex/Cursor, and an open-source skills repo as the adoption wedge. The founders' framing: 20 years of agency playbooks "codified into AI agents."

Two structural implications:
1. The connector layer is commoditizing. Channel integrations that used to justify hand-built plumbing now cost $9/month — a factory should buy its SENSE/ACT wiring, not build it.
2. Skills-as-distribution. Publishing the workflow (open-source skills, ClawHub agents) is the marketing — the same pattern as Larry's shared skill file. The playbook becomes both the product and the growth asset.

Notable restraint signal: the commercially shipping form sells predefined skills over live data — workflows, not open-ended autonomy — validating workflow-vs-agent at product level.

How It Applies to Marketing Factory

This is the tool-selection doctrine for the factory's stack: prefer channels and vendors that expose agent-drivable interfaces (MCP, clean APIs, webhook events) over dashboard-first tools, because every dashboard-only tool re-inserts a human into a loop the factory is trying to automate. It also reframes the gtm-engineering role — less integration plumbing, more skill/playbook authorship — and sharpens the factory's own moat question: if connectors and skills are commoditizing, the defensible layer is first-party data and the learned experiment history, not the wiring.

Referenced from: cogny-solo-platform