Which AI agents work with xSpeed?
24 agents have a dedicated guide. Each card shows how that agent signs in; open the guide for the exact steps, example prompts and a page for each job.
By maker
OpenAI
3 products- ChatGPT Chat app OpenAI's chat assistant. Developer mode on the web adds the Hub as an MCP app.
- ChatGPT dots Always-on agent OpenAI's always-on agents in ChatGPT. They use the Hub app you create in developer mode.
- Codex Terminal OpenAI's coding agent. codex mcp add takes the URL, codex mcp login signs you in.
Anthropic
3 productsxAI
3 products- Grok Chat app xAI's chat assistant. Add the Hub as a custom connector on grok.com and sign in once.
- Grok Bot Always-on agent xAI's named Bots on a shared cloud computer. Ask one to add the Hub, then attach it.
- Grok Build Terminal xAI's coding agent CLI. One grok mcp add command, then a browser sign-in.
By kind of client
Code editors
5 clients- Cursor Editor · Anysphere The AI code editor. One mcp.json entry adds the Hub; OAuth signs you in.
- GitHub Copilot in VS Code Editor · GitHub / Microsoft Copilot Chat in VS Code. One mcp.json entry; VS Code runs the sign-in.
- Windsurf Editor · Cognition Now Devin Desktop. Devin Local takes devin mcp add; legacy Cascade takes serverUrl.
- Zed Editor · Zed Industries A fast editor with an Agent Panel. Add the Hub under context_servers in settings.json.
- Kiro Editor · Amazon (AWS) The AWS agentic IDE. One mcp.json entry adds the Hub; Kiro opens the sign-in.
Autonomous agents
1 clientTerminal agents
2 clientsSelf-hosted agents
4 clients- OpenClaw Always-on agent · Open source The open-source assistant you chat with. mcp add, mcp login, and it reads your caches.
- Hermes Agent Always-on agent · Nous Research The self-hosted agent from Nous Research. A config.yaml entry, then hermes mcp login.
- Paperclip Always-on agent · Paperclip A self-hosted orchestrator for agent teams. Give an agent the Hub, then assign it tasks.
- NanoClaw Always-on agent · NanoClaw A small self-hosted assistant you own. Add the Hub to an agent group, then ask from chat.
Compare all 24 in one table
| Agent | Made by | Type | How it connects | Sign-in | Guide |
|---|---|---|---|---|---|
| Claude Code | Anthropic | Terminal | claude mcp add, then /mcp to sign in | OAuth sign-in or connection token | Claude Code guide |
| Claude | Anthropic | Chat app | Customize, Connectors, Add custom connector | OAuth sign-in or connection token | Claude guide |
| Claude Cowork | Anthropic | Desktop agent | A custom connector added to your Claude account | OAuth sign-in | Claude Cowork guide |
| ChatGPT | OpenAI | Chat app | Developer mode, then a custom MCP app | OAuth sign-in | ChatGPT guide |
| Codex | OpenAI | Terminal | codex mcp add --url, then codex mcp login | OAuth sign-in or connection token | Codex guide |
| Cursor | Anysphere | Editor | mcp.json with a url, then OAuth | OAuth sign-in or connection token | Cursor guide |
| GitHub Copilot in VS Code | GitHub / Microsoft | Editor | mcp.json in your VS Code user profile, or MCP: Add Server | OAuth sign-in or connection token | GitHub Copilot in VS Code guide |
| Windsurf | Cognition | Editor | devin mcp add, then devin mcp login (Cascade: mcp_config.json) | OAuth sign-in or connection token | Windsurf guide |
| Gemini CLI | Terminal | gemini mcp add, then /mcp auth xspeedhub | OAuth sign-in or connection token | Gemini CLI guide | |
| Antigravity | Editor | mcp_config.json with serverUrl, then Authenticate in settings | OAuth sign-in or connection token | Antigravity guide | |
| Zed | Zed Industries | Editor | context_servers in settings.json, or Settings, AI, MCP Servers | OAuth sign-in or connection token | Zed guide |
| Kiro | Amazon (AWS) | Editor | an mcp.json entry with a url, then a browser sign-in | OAuth sign-in or connection token | Kiro guide |
| OpenCode | SST | Terminal | a remote mcp entry in the config, then opencode mcp auth | OAuth sign-in or connection token | OpenCode guide |
| OpenClaw | Open source | Always-on agent | openclaw mcp add with --auth oauth, then openclaw mcp login | OAuth sign-in or connection token | OpenClaw guide |
| Hermes Agent | Nous Research | Always-on agent | an mcp_servers entry in config.yaml, then hermes mcp login | OAuth sign-in or connection token | Hermes Agent guide |
| Grok Build | xAI | Terminal | grok mcp add --transport http, then a browser sign-in | OAuth sign-in or connection token | Grok Build guide |
| ChatGPT dots | OpenAI | Always-on agent | the xspeedhub app from ChatGPT developer mode, shared through Plugins | OAuth sign-in | ChatGPT dots guide |
| Grok | xAI | Chat app | grok.com/connectors, New Connector, Custom | OAuth sign-in | Grok guide |
| Grok Bot | xAI | Always-on agent | a custom remote HTTPS MCP server plugin, added from chat | OAuth sign-in | Grok Bot guide |
| Muse | Meta | Chat app | a custom connector that Muse creates when you ask for it | Connection token | Muse guide |
| Manus | Manus | Always-on agent | Settings, Integrations, Custom MCP Servers, Add Server | Connection token | Manus guide |
| Kimi Code | Moonshot AI | Terminal | an mcpServers entry in mcp.json, then /mcp-config login | OAuth sign-in or connection token | Kimi Code guide |
| Paperclip | Paperclip | Always-on agent | claude mcp add on the agent's host, or Connect your own MCP server | OAuth sign-in or connection token | Paperclip guide |
| NanoClaw | NanoClaw | Always-on agent | ncl groups config add-mcp-server, then ncl groups restart | Connection token | NanoClaw guide |
Using a client that is not listed? Any MCP client that can reach a remote HTTP server works with the same URL. Agents that run shell commands on the server can also use WP-CLI.
How do I connect an AI agent to xSpeed?
4 steps, about two minutes. Each agent guide repeats them with that client's exact command or settings screen.
1Put your sites in xSpeed Hub
Sign in at app.xspeedcache.com with Google or email; the Hub is free and has no site cap. Then connect each WordPress site from its own dashboard: click Connect Hub in the xSpeed Cache top bar, then Connect via xSpeed Hub. Each site needs the free xSpeed Cache plugin.
# In each site's wp-admin xSpeed Cache → Connect Hub → Connect via xSpeed Hub
2Add one URL to your AI client
Every client on this page takes the same endpoint. Clients that sign in by URL open a browser page where you approve access. Clients that take headers can use the connection token from the Hub's Connect AI page instead.
# OAuth: paste the URL, approve in the browser https://app.xspeedcache.com/xspeed/mcp # Or with the connection token (Hub → Connect AI) Authorization: Bearer <your connection token>
3Decide what the agent may change
xSpeed Hub has no confirmation step of its own, so set this before your first request. Keep write tools on your client's approval prompt (each agent guide shows how), or connect read-only: the connection token with Read-only everywhere on, or a Viewer sign-in. A read-only connection can still call contact_support, which emails xSpeed support.
4Ask for what you want
Name the site, or ask about every site. The agent lists your sites first, then calls the tools it needs and reports what came back.
You: Which of my sites has the lowest cache hit ratio?
Purge that one and warm it again.
Agent: list_sites → 6 sites
get_cache_status × 6 → shop: 41% hit ratio
Purge and re-warm shop? (yes)
purge_cache shop · start_preloader shop What can an AI agent do with xSpeed?
10 jobs, each with a guide for every agent. Pick the job, then your agent.
Purge the cache
Clear stale pages after a deploy, an edit or a plugin update, on one site or all of them.
list_sitesget_cache_statuspurge_cachepurge_cloudflarestart_preloader
Find out why pages are not cached
Work out why a site misses the cache, serves slow pages or shows a low hit ratio, using read-only checks first.
list_sitesget_cache_statusget_healthrun_benchmarkget_benchmark_historyget_pro_audittoggle_cachecontact_support
Scan a site's speed
Get a graded speed report with the fix for every failing check and a link you can share, for your own site or any public URL.
list_sitesget_speed_scanrun_speed_scan
Track PageSpeed scores
Read the PageSpeed scores a site already has, run a new test when something changed, and tell a Lighthouse score from a cache benchmark.
list_sitesget_score_historyrun_speed_testget_benchmark_historyrun_benchmark
Raise the PageSpeed score
Preview and apply xSpeed's recommended settings to a site, check the pages still work, and be told plainly what caching cannot fix.
list_sitesget_score_historylist_modulesoptimize_sitegenerate_critical_csspurge_cacherun_speed_test
Tune cache settings
See which modules a site has, read their current settings, change them, and turn page caching on or off.
list_sitesget_cache_statuslist_modulesget_settingsupdate_settingstoggle_cacherun_benchmark
Warm the cache
Fill the page cache before visitors arrive, for example right after a purge or a deploy.
list_sitesget_cache_statusget_settingsupdate_settingsstart_preloaderget_preloader_statusstop_preloader
Set up Redis object cache
Connect a site to Redis or Memcached so database results survive between requests.
get_object_cache_statustest_object_cacheget_settingsupdate_settingstoggle_object_cacheflush_object_cache
Manage Cloudflare
Check the Cloudflare connection, clear the edge cache and switch development mode on or off for a site.
cloudflare_verifypurge_cloudflareset_cloudflare_dev_modepurge_cacheget_settings
How do AI agents connect to xSpeed?
Four surfaces reach the same plugin. Most people want the Hub: one URL in the AI client, every site behind it.
xSpeed Hub MCPMost people
https://app.xspeedcache.com/xspeed/mcp - Best for
- Most people. One URL in your AI client, every site behind it.
- Reaches
- Every site in your workspace, one connection
- Sign-in
- OAuth sign-in in the browser, or a connection token in an Authorization header
- Tools
- 29 tools (16 read, 13 write), seven of them can act on every site at once
xSpeed Cache Site MCP
https://your-site.com/xspeed/mcp - Best for
- A single site, or work the Hub does not expose, such as database cleanup.
- Reaches
- One WordPress site
- Sign-in
- OAuth sign-in by a site admin, or the site's own token in an Authorization header
- Tools
- The full per-site tool set plus run_command for the WP-CLI surface
xSpeed Scan MCP
https://xspeedcache.com/scan/mcp - Best for
- Grading a site you do not run, or trying xSpeed before you install anything.
- Reaches
- Any public URL, read-only
- Sign-in
- None. No account.
- Tools
- 5 tools: run_speed_scan, get_speed_scan and three product-info tools
WP-CLI
wp xspeed … - Best for
- Agents that already run commands on the server, such as Claude Code or Codex over SSH.
- Reaches
- The site whose server the agent has a shell on
- Sign-in
- Your server login
- Tools
- Every xSpeed command: cache, purge, preloader, objcache, cf, score, settings and more
The 29 tools an agent gets
Every tool in the xSpeed Hub MCP server. Reads change nothing on your sites; contact_support is the one read that acts, by emailing xSpeed support. Writes change a live site, or spend the shared speed-test allowance in the case of the two speed tests.
Write tools (13)
| Tool | What it does |
|---|---|
purge_cache writeevery site | Purge all, page, assets, object or REST cache. |
toggle_cache writeevery site | Turn page caching on or off. |
update_settings writeevery site | Change one module's settings. |
start_preloader writeevery site | Start warming the cache. |
stop_preloader writeevery site | Stop the cache preloader. |
optimize_site writeevery site | Apply a safe, standard or aggressive optimization pass, or preview it with dry_run. |
toggle_object_cache write | Turn the Redis or Memcached object cache on or off. |
flush_object_cache write | Empty the object cache. |
purge_cloudflare write | Purge the Cloudflare edge cache for a site. |
set_cloudflare_dev_mode write | Turn Cloudflare development mode on or off (about three hours). |
generate_critical_css write | Generate critical CSS. Needs the Pro Critical CSS module on the site. |
run_speed_test write, quota | Run a PageSpeed Insights test from the Hub, mobile or desktop. |
run_speed_scan write, quota | Run a full graded xSpeed Scan with a shareable report link. |
Read tools (16)
| Tool | What it does |
|---|---|
list_sites read | List the sites this connection can reach, with their handles and status. |
list_workspaces read | List your workspaces and which ones this connection reaches. |
contact_support read, sends email | Email xSpeed support on your behalf, with your account email and site list. The Hub classes it read because it changes no site, so a read-only connection can still send it. |
get_cache_status read | Cache on or off, stats and server details for a site. |
list_modules read | Free and Pro modules, their settings schema and status. |
get_settings read | Current settings for one module. |
get_health read | Full diagnostics for one site. |
get_pro_audit read | Which Pro features would help this site. |
get_benchmark_history read | Past cached versus uncached response-time runs. |
get_score_history read | Stored PageSpeed and GTmetrix history. Starts nothing. |
get_preloader_status read | Where the cache preloader is and what it has warmed. |
get_object_cache_status read | Redis or Memcached status, or the wp-config snippet. |
test_object_cache read | Live probe of the Redis or Memcached server. |
cloudflare_verify read | Check the Cloudflare connection on a site. |
get_speed_scan read | Read an xSpeed Scan report or a site's scan history. |
run_benchmark readevery site | Measure the site's own cached versus uncached response time. No Lighthouse score. |
Exact arguments: your client's tool list for this connection. Most tools share a name with a site MCP tool, but site-scoped Hub tools take a site argument and the schemas can differ. Access per site and per tool: scoped connections. Phrasing a request so the agent picks the right server and site: writing prompts for xSpeed Hub.
Can an AI agent change my site without asking?
How the Hub treats writes
xSpeed Hub has no confirmation step of its own. Read tools change nothing on your sites, with one exception to know about: contact_support sends an email to xSpeed support, and a read-only connection can still call it. A write tool (a purge, a settings change, a cache toggle, or a speed test that spends the shared allowance) runs as soon as your connection allows writes. Three things stand between a prompt and a change: your connection's scope, because a read-only connection refuses every write tool (the connection token has a Read-only everywhere switch, and a Viewer member only ever gets read access); your AI client's own tool-approval prompt; and the Hub's instructions to the agent, which tell it to confirm the target site and to ask once before a write that touches every site.
Your connection's scope
A read-only connection refuses every write tool, though it can still call contact_support, which emails xSpeed support. The owner can switch the connection token to Read-only everywhere on the Hub's Connect AI page; an OAuth sign-in carries the scopes the client asks for, and a Viewer member only ever gets read access.
Your client's approval prompt
Most clients ask before they call a tool unless you allowed it. Each agent guide says how that client asks and how to keep writes on the prompt.
The Hub's instructions to the agent
The Hub tells the agent to confirm an unclear site, to confirm setting changes and toggles, and to ask once before a write that touches every site. That is guidance to the model, not a lock.