ModelSaid

MCP server

Ask your assistant what AI says about you

Connect ModelSaid to the assistant you already use, and your visibility data is just there: scores, competitors, what changed this week. No dashboard, no copying numbers into a chat window.

Your API token

Paste a token and every snippet below fills in with it, ready to copy. This happens in your browser: the token is never sent to us, never put in the address bar, and never saved. Create one in your dashboard, under API tokens.

Snippets show a placeholder until you paste a token.

Where Run this in your terminal

claude mcp add --transport http modelsaid https://api.modelsaid.com/mcp \
  --header "Authorization: Bearer msk_your_token_here"

Add --scope user to make it available in every project rather than just this one.

Official docs: Claude Code

What your assistant can do

This table is generated from the same list the server enforces, so it cannot show you a tool that does not exist or put one on the wrong plan.

ToolWhat it doesPlanScope
list_projectsList the businesses in your account with their latest visibility score.Monitorprojects:read
get_projectGet one business: its details, latest score and when it is next scanned.Monitorprojects:read
compare_projectsCompare visibility scores across every business in the account.Monitorprojects:read
create_projectWritesCreate a project (a business to track). An account with an active subscription gets its first scan started immediately.Monitorscans:write
get_account_usageWhat the account has used of its plan this month: projects, on-demand scans and Auto Fix runs, with a per-project breakdown.Monitorprojects:read
get_project_slotsHow many project slots the account pays for, and how many are in use.Monitorprojects:read
get_latest_reportThe most recent visibility report for a business, including score components.Monitorreports:read
list_competitorsCompetitors the models named instead of, or alongside, this business.Monitorreports:read
get_trendScore history over time, bounded by the plan trend retention.Monitorreports:read
export_reportA full report rendered as Markdown, ready to paste into a deliverable.Monitorreports:read
list_model_answersThe verbatim answers each model gave, per prompt. The underlying evidence.Monitorreports:read
get_accuracy_findingsFactual inaccuracies the fact-check found in model answers about this business.Monitorreports:read
get_factsThe ground-truth facts the accuracy check measures model answers against.Monitorreports:read
update_factsWritesCorrect ground-truth facts. Only the fields you name change; the rest are kept.Monitorconfig:write
get_unified_reportOne project measured, in one call: latest report, snapshot, score trend and share of voice.Monitorreports:read
get_readinessThe technical AI-readiness scorecard for a business website.Monitorreports:read
list_changesRecent monitoring changes and alerts for a business.Monitorreports:read
list_technical_historyPast technical checks for a business: readiness, PageSpeed and availability.Monitorreports:read
list_issuesOpen issues for a business, ranked by severity decayed over recency.Monitorreports:read
get_monitoring_configWhether visibility and technical monitoring are on, how often, and the plan caps.Monitorprojects:read
set_monitoring_configWritesTurn visibility or technical monitoring on or off, and set cadence and extra pages.Monitorconfig:write
list_alert_rulesThe custom alert rules on a business, with the plan cap and how many are used.Monitorprojects:read
create_alert_ruleWritesAdd a custom alert rule watching one metric against a threshold.Monitorconfig:write
update_alert_ruleWritesChange an existing alert rule. Only the fields you name change.Monitorconfig:write
delete_alert_ruleWritesRemove a custom alert rule.Monitorconfig:write
list_alert_channelsWhere this account delivers alerts. Requires a token covering all projects.Monitorprojects:read
create_alert_channelWritesAdd an email or https webhook destination for alerts. Requires a token covering all projects.Monitorshare:write
delete_alert_channelWritesRemove an alert destination. Requires a token covering all projects.Monitorconfig:write
list_content_briefsContent opportunities computed from the latest scan.Monitorcontent:read
get_content_draftOne content draft with its AEO score and guardrail flags.Monitorcontent:read
create_content_draftWritesGenerate a content draft. Consumes one draft from the monthly quota.Monitorcontent:write
get_brand_profileThe brand guardrails drafts are generated and validated under.Monitorcontent:read
set_brand_profileWritesUpdate the brand guardrails profile. Only the fields you name change.Monitorconfig:write
refresh_content_briefsWritesRecompute content opportunities from the latest finished scan. No AI spend.Monitorcontent:write
dismiss_content_briefWritesDismiss a content opportunity. No later refresh brings it back.Monitorcontent:write
list_content_draftsThe drafts for a business, newest first, with the monthly quota used and its cap.Monitorcontent:read
update_content_draftWritesEdit a draft title or body, or approve a flagged draft. Edits are re-checked against the guardrails.Monitorcontent:write
run_scanWritesStart a visibility scan. Consumes one on-demand scan from the monthly quota.Monitorscans:write
list_scansScan history for a project: every scan with its status, score and timing.Monitorreports:read
get_scanOne finished scan: status, score and score components. Omit scan_id for the most recent.Monitorreports:read
run_auto_fixWritesPropose AEO fixes as a pull request on the connected repository. Never writes the default branch. Blocks for several minutes.Agency Proautofix:write
analyze_repositoryWritesInvestigate the connected repository and return file-level AEO recommendations. SPENDS MONEY: roughly USD 0.03 to 0.75 of the account's shared monthly AI budget, and one Auto Fix run against its daily and monthly allowance, unless a cached investigation of the same commit makes the call free. Two calls, always: the first one omits confirm_token, spends nothing, and returns the price and what the account has left for you to put to your human before you call again.Agency Proautofix:write
list_custom_promptsThe custom prompts this business is measured on, with the enabled cap and usage.Monitorprojects:read
create_custom_promptWritesAdd a custom prompt for the next scans to ask. Starts enabled, within the plan cap.Monitorconfig:write
update_custom_promptWritesChange or enable a custom prompt. Only the fields you name change.Monitorconfig:write
delete_custom_promptWritesRemove a custom prompt. Answers already collected with it are kept.Monitorconfig:write
get_prompt_suggestionsPrompts worth tracking, mined from Search Console queries and industry templates.Monitorreports:read
get_share_of_voiceOwn versus competitor share of AI answer mentions over time, in a date window.Monitorreports:read
get_roiThe stored revenue-at-stake assumptions and the estimate they produce.Monitorprojects:read
set_roi_inputsWritesReplace the revenue-at-stake assumptions. All fields at once, as on the dashboard.Monitorconfig:write
get_google_connectionWhether Search Console and GA4 are connected for a project, and which properties are selected.Monitorprojects:read
list_google_propertiesThe Search Console sites and GA4 properties the connected Google account can see.Monitorprojects:read
select_google_propertiesWritesChoose which Search Console site and GA4 property feed this project. Only the ones you name change.Monitorconfig:write
disconnect_googleWritesDisconnect Google from a project, deleting the stored tokens and the synced daily rows.Monitorconfig:write
get_search_console_dailyDaily Search Console clicks and impressions per query, from the connected site.Monitorreports:read
get_ga4_dailyDaily GA4 sessions and conversions per channel, from the connected property.Monitorreports:read
get_github_connectionWhether this project has a repository bound, and how many GitHub App installations the account holds.Monitorprojects:read
list_github_reposThe repositories every GitHub App installation on this account can reach.Agency Proprojects:read
select_github_repoWritesBind one repository to this project. Auto Fix runs read and target it.Agency Proconfig:write
list_github_branchesThe branches of the repository bound to this project, and the current base branch.Agency Proprojects:read
set_github_base_branchWritesChoose the branch Auto Fix runs are based on. Verified against the repository before it is stored.Agency Proconfig:write
disconnect_github_repoWritesUnbind this project from its repository. Other projects and the App installation are untouched.Agency Proconfig:write
create_share_linkWritesPublish one finished scan report at a public link that anyone holding the URL can read without signing in.Monitorshare:write
list_share_linksThe share links this account has published, with their public URLs, status, and how many active links each project has left.Monitorprojects:read
revoke_share_linkWritesRevoke a share link so its public page and badge stop resolving.Monitorconfig:write
get_white_label_profileThe white-label branding for shared and exported reports, and whether the plan includes it. Requires a token covering all projects.Monitorprojects:read
update_white_label_profileWritesChange the white-label branding on shared and exported reports. Only the fields you name change. Requires a token covering all projects.Monitorconfig:write

Available on Monitor, Agency, Agency Pro. Your assistant only sees the tools your plan and your token allow, so there is nothing to configure and nothing to accidentally call.

What a token can and cannot do

You choose the scopes
A token only carries what you tick. A read-only token cannot start a scan or open a pull request, whatever the assistant decides to try.
Your plan is checked every call
Not stored in the token. If a subscription lapses, the paid tools stop working on the very next request rather than whenever a token expires.
Every call is recorded
Which token, which tool, which project, and what happened. Arguments are not stored, so we do not keep a second copy of your business data.
Revoking is immediate
Revoke a token and the next call fails. There is no cached session to wait out and no refresh token still in flight.

The MCP server is on every paid plan

Monitor gets your own projects and reports. Agency adds work across clients. Agency Pro adds the raw model answers and lets an assistant open fix pull requests for you.