Security pluginsFree + Pro
Wordfence MCP: read findings, block addresses, harden settings
The Wordfence MCP abilities read reported firewall mode, scan issues, blocks and lockouts in GoBuildWP Free. Pro adds hardening, IP blocks and unblocks, and scan requests through Wordfence's save paths. Supported setting and block changes have recovery; scans have side effects and no undo. Range guards and loopback probes limit selected hazards, while proxy attribution and public login/API access require separate checks. Review the Wordfence Application Password setting before relying on that credential route.
The same /24 network keeps hitting wp-login.php. Block it for a day.
wppilot/security-lockoutsReadShows repeated login lockouts from that network over the last hours, with masked usernames.
wppilot/security-block-ipConfirm firstWith confirm: true, adds a Wordfence pattern block on the range for 24 hours with the reason given.
wppilot/security-lockoutsReadLists the new block with its expiry time.
- Edition
- Free 3 + Pro 5
- Abilities
- 8
- Requires
- WordPress 6.9+, PHP 8.0+
Abilities
The security abilities against Wordfence
| Ability | What it does here | Undo and limits |
|---|---|---|
security-plugin-statusFree | Wordfence's version, firewall on or off with its mode and protection level, scan schedule and last scan result, finding counts by severity, brute-force and two-factor settings, active lockouts and whether Premium is in use. | Read-only. The licence key is only tested for being set, never read out. |
security-scan-findingssecurity-lockoutsFree | Open scan issues, most severe first, naming the file relative to the site root or the plugin or theme slug; current IP blocks, pattern blocks and login lockouts with reason and expiry. | Read-only. Addresses come back as /24 or /48 networks and usernames in lockout reasons are replaced. |
security-harden-plansecurity-apply-settingsPro | The plan compares each Wordfence setting with the recommended value and gives a reason; apply recomputes it, writes each value through Wordfence, reads it back and checks the site and MCP routes before and after. | Needs confirm and the plan_id the person saw. Undo restores only settings still holding the value written; Application Passwords are never turned back off. |
security-block-ipsecurity-unblock-ipPro | Block adds an IP block, or a pattern block for a range, for duration_hours (24 by default, 0 for permanent) with the reason shown in Wordfence. Unblock lifts every block and login lockout on exactly that address, or the pattern block on exactly that range. | Confirmed and undoable: undoing an unblock recreates each block with its reason and remaining time. |
security-run-scanPro | Starts Wordfence's scan in the background; findings are read later. | Refused up front when Wordfence has no key from its onboarding. A scan changes no setting, so there is nothing to undo. |
Reference: /docs/reference/security
Shared module across these plugins; do not add this count across them. Edition and effect labels describe the source catalogue. The installed schema, dependency versions, licence, WordPress capability, safety profile and per-ability policy determine availability; labels alone do not prove confirmation or recovery.
What Wordfence records
Hardening Wordfence without locking out the agent
The plan turns learning mode into an enabled firewall, locks out after five failed logins for at least four hours, requires strong passwords and schedules scans, and it keeps any stricter value you already set. Blocking the "admin" username is offered only when no account has it. Two-factor for administrators is proposed only when the connected administrator has enrolled, so that setting does not lock out the person approving it. Other administrators, other sign-in routes and a proxy that hides real addresses are not covered by that check, so keep another way into the site before applying the plan.
Wordfence switches WordPress Application Passwords off by default, which makes every MCP call that signs in with one fail with a bare 401. The plan switches them back on, and neither the plan nor an undo ever turns them off. Applying writes each setting through Wordfence's own options path, reads it back, and checks the home page, the REST index and every MCP route before and after; if one that answered stops answering, the call attempts to restore the changed settings. Inspect rolled_back or rollback_failed and verify access; restoration can fail.
A block refuses the address the request came from, the server's own addresses, private and reserved ranges, ranges wider than /16 (IPv4) or /32 (IPv6), and an address Wordfence already blocks. Results name only the /24 or /48 network, lockout reasons have the attempted username replaced by a user id or masked form, and no licence or API key is read.
Limits
What stays with Wordfence and its owner
- Use a synthetic block target and retain independent administrator and file access. Recognized request, server and private ranges are refused; proxy attribution still needs verification. Check public login, API and critical routes after a change. A scan is requested or queued until a vendor result is read.
- Extended protection is reported, not set up: Wordfence's Optimize step edits server configuration, which GoBuildWP does not write.
- The Premium real-time IP blocklist and country blocking are not managed here.
- The scan ability refuses a missing onboarding key. Whether an installed Wordfence scan completes depends on its connectivity, background processing and vendor requirements; inspect its scan log on local or private sites.
- Hiding the login URL, disabling XML-RPC or the REST API for signed-in users and blocking this request's own address are never offered.
Worked example
Stopping a range that keeps probing the login page
Expected outcome to verify on the installed stack; this sequence has not been recorded as a runtime test.
Wordfence blocks the range until tomorrow and then drops the block itself. The block can be undone from the change log before then. The call would have been refused if the range covered the address the request came from or the server's own addresses.
Review the exact host or range and collateral access before blocking. Masked request logs cannot identify an exact safe range. Check ordinary public, login and API routes after the change.
security-lockoutsShows repeated login lockouts from that network over the last hours, with masked usernames.
security-block-ipWith confirm: true, adds a Wordfence pattern block on the range for 24 hours with the reason given.
security-lockoutsLists the new block with its expiry time.
Data model
Wordfence's config table, issues and blocks
Wordfence keeps its settings in its own configuration store, read and written through wfConfig, and validates every value before it saves it; the abilities use that same validate-then-save route. Firewall state comes from Wordfence's firewall object: whether it is enabled, in learning mode or disabled, and at which protection level.
Scan results are rows in Wordfence's issues table, and only those still marked new are open findings. Blocks, pattern blocks and login lockouts are records in Wordfence's block list, which deletes them when they expire, so the lockout list shows what is in force now rather than a history.
Scan triage, a scan that did not finish and block lengths
Turning a Wordfence scan report into a short list of jobs
Most Wordfence MCP sessions begin with security-scan-findings, filtered to critical and high. Each open issue comes back with its severity, Wordfence's issue type, a short description and either a file path relative to the site root or, for plugin and theme issues, a component with its type and slug. That last field is what makes the list actionable: an outdated or vulnerable plugin finding names the exact slug the update abilities need, so the agent can propose the update and run it through GoBuildWP's safe update flow instead of leaving the finding open.
Findings about modified core files or unknown files are different. The agent can name the file and the issue type, but it does not repair or delete files through this module; compare and repair them in Wordfence's own scan screen, where Wordfence shows the difference against the original. Issues an owner has ignored in Wordfence are left out of the list, so an empty result after ignoring is not the same as a clean site.
When security-run-scan started but no result arrives
Wordfence scans in the background, so security-run-scan returns started and the agent has to come back for the result. security-plugin-status shows when the last scan finished and its result, and when a scan failed, the failure type Wordfence recorded. A scan that never moves past started can mean the host is not running background tasks, or that Wordfence cannot reach the site, as with a local copy; both are things to check in Wordfence's scan log rather than to retry blindly.
The run is refused immediately when Wordfence has no key from its onboarding, with the fix named, and also while a scan is already running. A free key from Wordfence's onboarding is enough.
How long a Wordfence block should last
security-block-ip takes duration_hours, 24 by default, and 0 for permanent. For a network hammering wp-login.php a day is usually enough, and Wordfence removes the block by itself when it expires. Because Wordfence deletes expired blocks, the lockout list shows what is in force now, not a history; the reason you give is the only record of why a block existed, so make it specific. Permanent blocks on residential ranges tend to catch real customers later, after addresses are reassigned.
The three read abilities are part of GoBuildWP Free and work on the Read Only profile. Blocking, unblocking, scanning and applying a plan are Pro, and all but the scan need confirm on Production Safe. The Solid Security page covers the same abilities against Solid.
- "Read only: list critical and high Wordfence findings and group the plugin ones by slug."
- "Start a Wordfence scan and tell me when it has finished and what changed in the findings."
- "Block the network that keeps failing logins for 48 hours, with the reason 'login flood, October'."
Questions
Wordfence questions
Why do MCP calls fail with a 401 after installing Wordfence?
Wordfence switches WordPress Application Passwords off by default, and many MCP clients sign in with one. GoBuildWP Pro's hardening plan switches them back on, and an undo never turns them off again; on GoBuildWP Free, switch them back on in Wordfence's own settings or connect with OAuth.
Can an AI agent block its own IP in Wordfence?
The guard refuses recognized request and server addresses, private ranges and ranges wider than /16 or /32. Proxy attribution still needs verification; this does not prove that every route or administrator is protected from lockout.
Why does the agent say the Wordfence scan was not run?
Wordfence needs a key from its onboarding before it scans. Without one the ability refuses and names the missing key.
Can I undo a Wordfence setting the agent changed?
Yes, from the change log, as long as nobody has changed that setting since. A setting changed afterwards is named and left alone.
Where these facts come from
GoBuildWP behaviour is taken from the plugin source and the ability reference; third-party details from the vendor documentation below, checked by the GoBuildWP team.
Wordfence is a separate product; no partnership or endorsement is implied.