Integration · Frase

Read-write MCP across the SEO and GEO content lifecycle.

Frase's first-party MCP server is one of the few read-write MCPs in the SEO content optimization category. The agent does not just read scores; it writes briefs, drafts, edits, and publish actions back into the platform.

Six pipeline stages reachable from one MCP connection. Research: SERP analysis, keyword discovery, competitor outline mining. Brief generation: structured outlines with target keyword, intent, recommended word count, entities to cover, and dual SEO + GEO scoring targets. AI writing: generation calibrated to the team's brand voice. Optimization: real-time SEO + GEO scoring against the top-ranking SERPs. Publishing: direct push to Frase's CMS, WordPress, Webflow, or Sanity. Monitoring: AI visibility tracking across 8 platforms and Content Watchdog for autonomous ranking-recovery flags. MCP access ships included on all plans starting at $49 per month.

Official MCPFrase official MCP, API key auth, read-write across the full content lifecycle.

What you can do via MCP

Example prompts the agent runs.

  • Research the top 10 SERPs for 'experimentation analytics'; tell me the entities every page covers and the headings the top three share.

    Calls the SERP-analysis tool, returns the entity intersection across top 10 results, and the headings shared by top 3. The intersection is what a competitive brief has to cover; the divergence is where the brand can differentiate.

  • Generate a brief for 'experimentation analytics dashboards' targeting commercial-investigation intent; aim for 2,400 words.

    Creates a structured brief: target keyword, intent, word count, recommended H1 plus H2 outline, entities to cover, dual SEO + GEO score targets. Returns the brief object the writer or the agent picks up next.

  • Write the introduction for the brief I just generated; calibrate to the brand voice doc tagged 'rampstack-direct-prose'.

    Calls the AI writing tool with the brief context plus the brand-voice fingerprint, returns a calibrated draft of the intro. The voice fingerprint is the lever that keeps the output from reading generic.

  • Score this draft against the SERP target; tell me which entities I'm missing and where my GEO score is below 80.

    Returns the dual-score breakdown (SEO score plus GEO score), the missing entities, and the sections where each score lags the target. GEO is the AI-search-engine-readability score; the missing entities are the agent's prioritization signal.

  • Show me which articles dropped in AI search visibility this month and trigger Content Watchdog on the worst three.

    Calls the AI visibility monitoring tool across 8 platforms, returns articles with the largest visibility deltas, and triggers Content Watchdog on the worst three to flag autonomous ranking-recovery actions.

Frase · MCP

Brief generation via the Frase MCP. The agent specifies the target keyword and intent; Frase analyzes the SERP and returns a structured brief with dual SEO + GEO targets.

Frase MCP
docs
# Frase MCP, called via the agent runtime
mcp.frase.create_brief({
  target_keyword: "experimentation analytics dashboards",
  search_intent: "commercial-investigation",
  word_count_target: 2400,
  competitor_serp_count: 10,
  scoring_mode: ["seo", "geo"],
  brand_voice_id: "rampstack-direct-prose"
})

# Returns: brief_id, target_keyword, intent, word_count_target,
#          headings: [H1, H2 x N, H3 x M], entities, score_targets {seo, geo}.
One command sample showing how the agent talks to Frase. The MCP exposes the platform's primitives as tools; the agent translates the prompt into the right call.

MCP integration

Frase MCP server.

Server
Frase official MCP endpoint (see Frase docs for the latest URL)
Auth
API key (per-account, rotatable in Frase settings)
Hosting
Frase hosted
Scoping note
MCP access ships included on every Frase plan starting at $49 per month. Token scoping respects the account's role; admin tokens can write across all projects, editor tokens are scoped to assigned projects.
  • Read-write MCP: research, brief, write, score, publish, monitor, all reachable from one connection
  • Dual SEO + GEO scoring: traditional SEO ranking signals plus AI-search-engine readability
  • AI visibility monitoring across 8 platforms (ChatGPT, Perplexity, Gemini, Claude, and more)
  • Content Watchdog: autonomous ranking-recovery flags for articles whose visibility drops
  • Brand-voice calibration: AI writing matches the team's voice doc, not generic LLM output
  • Publishing connectors to Frase CMS, WordPress, Webflow, and Sanity

Frase docs

Visual demonstration

What this looks like in practice.

Frase· Content briefAI-generated outline
Target keywordexperimentation analytics dashboards
Intentcommercial-investigation
Word count2,400 words
SEO86/100

Top 10 SERP target reached

GEO72/100

AI-search readability, target 80

Recommended outline

  • H1Experimentation Analytics Dashboards: A 2026 Guide
  • H2What an experimentation analytics dashboard does
  • H3Required metrics: lift, p-value, confidence interval, MDE
  • H3Common drift modes that hide on default dashboards
  • H2Two operational models: platform-native vs warehouse-native
  • H3Platform-native: faster, narrower, opinionated
  • H3Warehouse-native: slower, broader, custom-metric capable
  • H2Picking the right model for the team's volume and depth
  • H2Build vs buy: when to graduate from platform to warehouseinclude cost table
  • H2Common dashboard patterns that ship trusted resultsship 3 examples

Entities to cover

  • experimentation
  • A/B testing
  • statistical significance
  • minimum detectable effect
  • CUPED
  • sequential testing
  • Statsig
  • PostHog
  • Optimizely
  • data warehouse
  • dbt
  • Snowflake
  • Mixpanel
The brief Frase produces from a target keyword and intent. H-tag outline plus dual SEO + GEO scores plus the entities the top 10 SERPs share. The agent uses this object as the structured input for the AI writing call that comes next.

CLI alternative

Frase REST API for non-MCP workflows.

Frase exposes a REST API alongside the MCP for embedded workflows: programmatic brief generation in CMS plugins, scheduled scoring runs in CI, and bulk operations across hundreds of articles at once. The MCP is the primary surface for AI-driven editing inside the agent runtime; the REST API fills in for the operational shape the agent runtime is not built for.

Pairs with these skills

The content stack skill suite.

This integration pairs with the forthcoming content-strategy-for-ai-search and content-production skills. The Frase microsite assumes the content strategy exists; the integration shows where the lifecycle runs. Skill pages and SKILL.md sources land in subsequent dispatches; cross-link hyperlinks are added when the skill pages ship.

On Surfer SEO, honestly.

Frase is one option in the SEO content optimization category. Surfer SEO is the most-recognized alternative. The honest comparison: Surfer has a broader install base among traditional SEO teams and stronger visual editors; Frase has the read-write MCP today and ships GEO scoring plus AI visibility monitoring inside the same surface. Surfer's MCP support is community-built and read-only at the time of writing; Frase's is first-party and read-write. For teams whose primary motion is agent-driven content production with AI search visibility as a tracked outcome, the Frase MCP is the closer fit. For teams whose motion is human-led editing with the agent as a research assistant, Surfer's editor still leads. Pick by motion, not by feature checklist.