Every site's data and tools, in one connection for your agent.
Sitearc's MCP server works with Claude, Claude Code, Codex and other MCP clients. Your agent gets 130 tools to read your sites, research, and file work for you to review.
What your agent gets
- 01
Every site, one connection
One connection covers every site you can open, across the workspaces in your account. Your agent names the site on each call.
- 02
Your own data, free to read
Search Console, GA4, audits, forms and feedback. Reading them never spends credits.
- 03
Research, paid with credits
Keywords, rankings, competitors, backlinks and AI visibility. Rank checks are priced before they run.
- 04
Work it can file
Your agent turns findings into ideas and proposals with evidence. It opens pull requests for you to review.
- 05
Brand and memory
It reads your brand kit, project context and past learnings. Each session starts from what you already know.
- 06
A split you control
61 tools only read. A read-only key sees only those, and cannot write, spend credits or publish.
All 130 tools, grouped by job.
61 only read. The other 69 may change something or spend credits, and none of them can approve.
- 9 tools7 readSites and the connectionWho is connected, which sites it reaches, and where to start a session.
whoamilist_projectscreate_project - 12 tools12 readSearch Console and AnalyticsRead your own search and visitor data. These never spend credits.
get_search_console_performanceinspect_urlsget_search_opportunities - 8 tools3 readSite health and crawlingAudit a site, read its issues, and fetch or crawl pages.
run_site_auditget_audit_statusget_audit_issues - 14 tools2 readRankings and local searchTrack keyword positions, and check how a business shows up in local results.
create_rank_trackerget_rank_trackeradd_rank_tracking_keywords - 13 tools1 readResearchKeywords, search results, competitors, backlinks and AI visibility. New lookups spend credits; saved keywords are free.
research_keywordsget_keyword_metricslist_saved_keywords - 11 tools5 readFeedback and formsOpen client reviews, read their comments, and set up and test forms.
create_reviewlist_review_commentsresolve_review_comment - 15 tools7 readWork queue and GitHubTurn ideas into proposals, open pull requests and submit them for your review.
list_recommendationsget_recommendationupdate_recommendation_status - 13 tools4 readContentBriefs and drafts your agent writes, sent to a person before anything is published.
create_content_brieflist_content_backlogget_content_document - 9 tools5 readExperiments and learningsRecord what shipped, measure it against holdout pages and keep what was learned.
ship_experimentlist_experimentsget_experiment - 9 tools6 readBrand kit and project memoryRead the brand kit and shared context, and propose changes to them.
get_brand_profileget_compliance_profilelist_brand_assets - 11 tools7 readSetting up new sitesPlan a new Astro site or Google measurement. A person approves each plan first.
discover_setup_sourcelist_setup_runsplan_astro_setup - 6 tools2 readWordPressThese need the Sitearc WordPress plugin, which is not released.
wp_list_postswp_get_postwp_update_post
Connect in a minute.
Bring the agent you already use. Sitearc runs no AI of its own, so there is no model markup.
Claude on the web
OAuth, no keys- 1In Claude, add a custom connector.
- 2Paste https://app.sitearc.ai/mcp as its URL.
- 3Choose Sign in now, then Register automatically if asked.
- 4Sign in to Sitearc and allow access.
Claude Code and Codex
One commandAdd the server once for every project. Then run /mcp in Claude Code to sign in.
$ claude mcp add --transport http sitearc --scope user https://app.sitearc.ai/mcpCodex takes the same address.
$ codex mcp add sitearc --url https://app.sitearc.ai/mcpAn API key
For scriptsCreate a key in Sitearc's settings. Keys start with oseo_.
x-api-key: oseo_…Or send it as a Bearer token. Choose read or write when you create it. A read key sees only the 61 read tools.
What no tool can do
- Approve. Proposals and previews are approved in the dashboard. A test in the app fails if an approve tool appears.
- Merge or deploy. Your agent opens the pull request. Merging happens in GitHub.
- Pick its own result. Sitearc computes won, backfired or inconclusive from the measurements.
What your agent can do
Your agent starts a session with tools like these. Some only read. Others may change something or spend credits.
get_Readsproject_ brief Starts a session with the site, its repo, its sources, its learnings and running experiments.
list_Readsprojects Lists every website your agent can reach.
whoamiReadsConfirms which account the agent is connected as, its scopes and the credit balance.
get_Readsbudget_ status Reads the spend limits that apply to a site, and what has been spent this period.
get_Readssearch_ console_ performance Reads clicks, impressions, CTR and position by query, page, country, device or date.
get_Readsaudit_ issues Reads the issues from the last audit by priority, each with how to fix it.
estimate_Readsrank_ tracker_ cost Prices a check or a schedule before anything is spent.
run_May change or spendrank_ tracker Runs a live check now, within a credit limit you approved.
How it works
- 1
Add app.sitearc.ai/mcp to Claude, Claude Code, Codex or another MCP client, and sign in once.
- 2
Your agent reads the project brief: the site, its sources, its learnings and what is running.
- 3
It researches, files ideas and proposals with evidence, and opens pull requests through GitHub.
- 4
You approve proposals and previews in Sitearc. Sitearc only records a change as shipped when what merged is exactly what you approved.
Limits and credits
Know the edges before you rely on it. Credits pay for outside data and heavy work, and saved data never costs any.
Limits
- No tool can approve a proposal or a preview. That happens in the dashboard, signed in.
- Sitearc never merges or deploys. It can't stop someone merging a pull request in GitHub.
- Free includes 1,000 MCP requests a month on saved data. Paid data tools need Pro credits.
- Claude on the web is tested end to end. Claude Code and Codex steps are documented, not tested on every client.
- The WordPress tools need the Sitearc WordPress plugin, which is not released.
- The Sitearc plugin and skills for Claude Code and Codex are not public yet.
Credits
Keyword search Per call; about 256 in auto mode (3 calls) | 80 credits |
|---|---|
SERP analysis At depth 10; about 67 at depth 100 | 7 credits |
Backlinks overview About 347 for a full lookup | 163 credits |
AI visibility lookup Measured; up to about 3,500 | 2,086 credits |
Rank check Per keyword, device and check, queued at depth 10; 20 at depth 100; live checks 7 to 67 | 2 credits |
One credit is $0.001. Pro includes 100,000 a month, and Free includes none. See pricing
Connect your first site free.
Add every site you run, at no cost. Your agent can start the same day.