Adform’s Solutions Engineering team released 29 agentic skills on GitHub that let AI agents query the Adform FLOW demand-side platform in plain language. The skills run through the Adform GraphQL Model Context Protocol server and cover reporting, forecasting, audience discovery, and campaign and creative audits, according to PPC Land. Every skill is read-only.

Read-Only by Design

Each skill is a single markdown file that defines its purpose, illustrative GraphQL queries, usage constraints, and presentation guidance. There is no executable code bundled with most of them and no write access to FLOW. An AI agent connected to the Adform GraphQL MCP server reads the skill file, learns which queries answer a given question, and returns a report. It cannot push changes back into the account.

The taxonomy governance skill, one of the 29, illustrates the pattern: it pulls campaign lists for an advertiser, applies a naming convention regex on the client side, and flags any campaign name that fails to match. A second pass checks label taxonomy against what has actually been applied. The output is prescribed as compliance tables leading with non-compliant counts. As the skill file states: “Read-only, it reports; the trafficker fixes.”

Rate limiting is built directly into the skill operating instructions, with calls to the GraphQL endpoint spaced one to two seconds apart.

The MCP Pattern in Ad Tech

Adform joins a growing list of ad tech platforms standardizing on MCP for agent access. Yahoo launched MCP integration in January 2026, and Meta followed with its connector in April, according to PPC Land. The pattern across all three: open RPC-style endpoints that let Claude, ChatGPT, and other agents query campaign data without granting write permissions.

Suren Silva, VP of Global Solutions Engineering at Adform, announced the release on LinkedIn, crediting colleagues Marcel Ehrlitzer, Hans Jirschik, and Maja Sokołowska along with the Product and Engineering teams. Silva noted that more skills are planned.

The 29 Skills

The GitHub repository covers bid landscape analysis, budget risk monitoring, campaign management and performance, channel conflict auditing, creative discrepancy auditing, deal health checks, delivery monitoring, frequency cap auditing, geo reference lookups, inventory availability forecasting, media plan review, pacing checks, reach forecasting, reporting, segment governance, stale entity auditing, taxonomy governance, tracking tags, viewability auditing, and audience discovery, among others. The scope suggests Adform is treating agent access as a first-class product surface rather than a side feature.

For marketing operations teams already using AI assistants for analysis, structured read-only access to DSP data removes the manual export step that currently sits between question and answer.