Connect your AI agent

Start here. Every website with its id, running work, latest audit, task counts and open questions, plus connections, provider quota and recent activity.

No parameters.

Wait until work starts, finishes, blocks or asks a question, or the time runs out. Returns status. Use between long-running tool calls instead of polling.

secondsinteger

Longest wait; default 30, at most 50.

The content calendar, past days included: per day how many tasks were assigned and how many got built, and the tasks themselves. A task missed on a day is listed there with movedTo. Plus tasks on no day and tasks needing attention.

siteIdrequiredstring

Website id from status.

fromstring

First day, YYYY-MM-DD in the Mac’s local time; default today. Use an earlier day to read history.

daysinteger

Days to return; default 14, at most 90.

What is happening now, as Today shows it. Omit siteId for every workspace, audit and pipeline readiness, combined counts and task lists. Pass siteId for one website’s build step, queue, pause, questions, attention, built pages and execution/audit models.

siteIdstring

Website id from status.

The app’s activity log, newest last: what started, finished and failed.

limitinteger

Entries; default 40, at most 300.

Connected websites whose public page templates can be reused. The same optional choices as Website settings; nothing is selected automatically. Select with site_save templateSiteId, or clear it with an empty string.

siteIdstring

Target website id from status; omit while adding a website.

A website’s full settings: contract directories, commands, agents, build setup (references, components, questions), decisions and day run.

siteIdrequiredstring

Website id from status.

First-run connection and website readiness, using the same checklist as the window.

No parameters.

The standards build agents follow. Without a name: the list. With a name: the built-in text and the owner’s additions.

namestring

Playbook name.

Connect a website or change its settings. Pass id to edit; then only the fields you send change. New websites need url and repoPath.

idstring

Existing website id. Omit to connect a new website.

siterequiredobject

Website settings.

site.namestring

Display name.

site.urlstring

Public website URL.

site.repoPathstring

Absolute path to the root of the website’s local Git repository.

site.templateSiteIdstring

Optional connected website id from setup_templates whose public templates the agent may reuse. Empty clears it. Never copies its product facts, credentials or publishing settings.

site.descriptionstring

What the product is.

site.audiencestring

Who it is for.

site.productFactsstring

Verified product facts agents may state.

site.articlesDirstring

Repository directory for articles.

site.toolsDirstring

Repository directory for free tools.

site.pagesDirstring

Repository directory for marketing pages.

site.repairDirsstring[]

Shared code directories agents may edit.

site.allowedSupportingFilesstring[]

Exact supporting files agents may edit.

site.toolApiDirstring

The one public endpoint directory free tools may add to, e.g. app/api/tools. Empty clears it.

site.validateCommandstring

Command that validates the website, e.g. npm run build.

site.publishCommandstring

Deploy command used by pipeline_ship, e.g. git push.

site.articlePublishCommandstring

For a blog stored in a database table: command that publishes one article draft file, e.g. node scripts/publish-article.mjs. The draft path is appended. Article tasks then write one draft in the articles directory and ship without a commit or deploy.

site.providerclaude | codex | opencode | ollama

Executor provider.

site.modelstring

Executor model; empty uses the provider default.

site.effortstring

Executor effort.

site.managerProviderstring

SEO audit provider; empty means same provider as execution.

site.managerModelstring

SEO audit model. Builds never use this model.

site.managerEffortstring

SEO audit reasoning effort.

site.publicationPolicyreview | auto

Publication policy.

site.monthlyBudgetUsdnumber

Monthly research budget for this website.

site.countryCodeinteger

DataForSEO location code, e.g. 2840 for the United States.

site.languageCodestring

Language code, e.g. en.

site.keywordsstring[]

Seed keywords.

site.visibilityPromptsstring[]

Buyer questions measured in AI answers.

Disconnect a website and delete its audits, tasks and measurements from ClicksPal. The repository is untouched.

idrequiredstring

Website id from status.

Re-read the repository to detect its framework and fill empty product facts.

siteIdrequiredstring

Website id from status.

Have the execution model read the repository and save reference pages, shared components, decisions and owner questions. Runs once before the first build; rerun after large repository changes. Returns as soon as the work starts. Call wait, then status, to follow it.

siteIdrequiredstring

Website id from status.

Stop a running website setup inspection and cancel builds waiting for it. Other websites, audits and completed work are untouched. Run setup_run or job_run again when ready.

siteIdrequiredstring

Website id from status.

Answer build-setup questions (site_get → buildSetup.questions).

siteIdrequiredstring

Website id from status.

answersrequiredobject[]

One answer per open question key.

answers[].keyrequiredstring

Question key.

answers[].answerrequiredstring

The answer; pick one of the offered options when they fit.

Choose the page later work of one content type starts from. An empty path clears it.

siteIdrequiredstring

Website id from status.

kindrequiredmarketing | tool | article

Content type.

pathrequiredstring

Page file path relative to the repository root.

Save a settled answer every later agent treats as authority for this website.

siteIdrequiredstring

Website id from status.

keyrequiredstring

Short identifier, e.g. tool-ai-provider.

questionrequiredstring

What was decided.

answerrequiredstring

The decision.

Delete a saved website decision.

siteIdrequiredstring

Website id from status.

keyrequiredstring

Decision key from site_get.

Save owner additions to a build standard; they win over the built-in text. Empty content removes them.

namerequiredstring

Playbook name from playbooks_get.

contentrequiredstring

Markdown additions, under 20,000 characters.

siteIdstring

Unused; standards apply to every website.

A website’s saved SEO audits, oldest first.

siteIdrequiredstring

Website id from status.

An audit’s overview: scores per dimension, the five conclusions, research progress per content type, organic baseline, competitors and limitations. Latest audit unless auditId is given.

siteIdrequiredstring

Website id from status.

auditIdstring

Audit id; omit for the latest.

includestring[]

Bulky sections to add.

An audit’s keyword opportunities with measured volume and difficulty, filtered and sorted. detail adds the brief, evidence and secondary keywords. Unknown volume is null, never zero.

siteIdrequiredstring

Website id from status.

auditIdstring

Audit id; omit for the latest.

lanemarketing | tool | alternatives | listicle | article | programmatic

Content type.

dispositionvalidated | low-data | rejected | all

Research verdict; default validated and low-data.

statussuggested | approved | deferred | rejected | producing

Owner decision.

fitdirect | adjacent | weak

Product fit.

searchstring

Text in the keyword, title or secondary keywords.

minVolumeinteger

Lowest monthly searches.

maxDifficultyinteger

Highest difficulty, 0–100.

sortpriority | volume | difficulty

Order; default priority.

idsstring[]

Return exactly these audit keyword ids.

detailboolean

Full records.

limitinteger

Rows to return; default 50, at most 500.

offsetinteger

Rows to skip.

An audit’s technical findings grouped by root cause, worst first. detail adds evidence, the repair brief and every affected URL.

siteIdrequiredstring

Website id from status.

auditIdstring

Audit id; omit for the latest.

severityerror | warning | info

Only this severity.

detailboolean

Full records.

limitinteger

Rows to return; default 50, at most 500.

offsetinteger

Rows to skip.

The website’s existing pages from an audit: type, indexability, inbound links and observed Search Console clicks and impressions. Null traffic means not observed, not zero.

siteIdrequiredstring

Website id from status.

auditIdstring

Audit id; omit for the latest.

lanemarketing | tool | alternatives | listicle | article | programmatic | other

Page type.

searchstring

Text in the URL or title.

limitinteger

Rows to return; default 50, at most 500.

offsetinteger

Rows to skip.

Workspace opportunities: the owner’s decisions on audit keywords plus ones added by hand or from Search Console and AI answers.

siteIdrequiredstring

Website id from status.

statussuggested | approved | deferred | rejected | producing

Decision.

sourcemanual | agent | dataforseo | audit | gsc | visibility

Origin.

formatarticle | alternatives | listicle | tool | marketing | programmatic | repair

Content type.

searchstring

Text in the keyword or title.

detailboolean

Do not shorten the rationale.

limitinteger

Rows to return; default 50, at most 500.

offsetinteger

Rows to skip.

The latest technical crawl: health score, checks, page speed, crawler access and its issues.

siteIdrequiredstring

Website id from status.

includeResolvedboolean

Include resolved issues.

includePagesboolean

Include every crawled page.

Run the full SEO audit: crawl, existing pages, and keyword research for every content type. Spends the saved DataForSEO budget; cached measurements are reused. Returns as soon as the work starts. Call wait, then status, to follow it.

siteIdrequiredstring

Website id from status.

agentobject

Model for this audit. Omit to use this website’s saved audit model, falling back to its execution model.

agent.providerrequiredclaude | codex | opencode | ollama

Provider; check status for remaining quota.

agent.modelstring

Model id from settings_get; empty for the account default.

agent.effortstring

Reasoning effort; empty for the default.

Stop the running SEO audit and keep the research finished so far.

siteIdrequiredstring

Website id from status.

Write the five overview conclusions for a finished audit. No-op when they already exist. Returns as soon as the work starts. Call wait, then status, to follow it.

siteIdrequiredstring

Website id from status.

auditIdrequiredstring

Audit id from audits_list.

Fetch the daily clicks and impressions chart for an audit. Returns as soon as the work starts. Call wait, then status, to follow it.

siteIdrequiredstring

Website id from status.

auditIdrequiredstring

Audit id from audits_list.

Turn the monthly audit rerun on or off.

siteIdrequiredstring

Website id from status.

enabledrequiredboolean

true reruns the audit every 30 days while the app is open.

Write an audit and its task list to a JSON file. Use this instead of paging audit_keywords when you need everything.

siteIdrequiredstring

Website id from status.

auditIdstring

Audit id; omit for the latest.

pathrequiredstring

Absolute .json path in an existing folder.

Mark a technical-crawl issue resolved or open.

issueIdrequiredstring

Issue id from crawl_get.

resolvedrequiredboolean

true resolves it.

Run only the technical crawl (older flow; audit_full includes it). Results are in crawl_get. Returns as soon as the work starts. Call wait, then status, to follow it.

siteIdrequiredstring

Website id from status.

Have an agent repair chosen technical-crawl issues in an isolated worktree. Returns as soon as the work starts. Call wait, then status, to follow it.

siteIdrequiredstring

Website id from status.

issueIdsrequiredstring[]

Issue ids from crawl_get.

Have an agent repair every open error and warning from the technical crawl. Returns as soon as the work starts. Call wait, then status, to follow it.

siteIdrequiredstring

Website id from status.

Add or edit a workspace opportunity. Demand numbers are never accepted from callers; research_run measures them.

siteIdrequiredstring

Website id from status.

idstring

Existing opportunity id. Omit to add.

opportunityrequiredobject

Opportunity fields.

opportunity.titlerequiredstring

Title.

opportunity.keywordrequiredstring

Target keyword.

opportunity.formatrequiredarticle | alternatives | listicle | tool | marketing | programmatic | repair

Content type.

opportunity.intentstring

Search intent.

opportunity.rationalestring

Why this is worth building.

opportunity.sourcemanual | gsc | visibility

Where it was observed.

opportunity.sourcesstring[]

Evidence URLs. For gsc and visibility they must come from this website’s saved report or AI answers.

opportunity.existingUrlstring

Existing page to improve instead of creating a new one.

Keep, defer or reject an opportunity. The calendar follows the decision.

idrequiredstring

opportunityId from audit_keywords, or id from opportunities_list.

statusrequiredsuggested | approved | deferred | rejected

Decision.

Create a task from a workspace opportunity and start building it. Returns as soon as the work starts. Call wait, then status, to follow it.

idrequiredstring

Opportunity id from opportunities_list.

Have an agent propose opportunities outside an audit, then measure them (older flow; audit_full is the main path). Spends the saved DataForSEO budget; cached measurements are reused. Returns as soon as the work starts. Call wait, then status, to follow it.

siteIdrequiredstring

Website id from status.

Measure search volume and difficulty for keywords. Spends the saved DataForSEO budget; cached measurements are reused. Returns as soon as the work starts. Call wait, then status, to follow it.

siteIdrequiredstring

Website id from status.

keywordsstring[]

Up to 200 keywords; omit to measure the website’s saved keywords.

The Content pipeline list for one day, as the window shows it: now (building), next (today: in the exact order Autopilot builds them, stopped tasks it retries first; another day: what is scheduled) and done (built or live that day). Questions and stopped builds are in status needsYou. Omit siteId for every workspace.

siteIdstring

Website id from status.

daystring

Day, YYYY-MM-DD in the Mac’s local time; default today.

What ClicksPal has published, as the Published pages screen shows it: every live page per workspace grouped by type, plus pages added and pages updated per day (new pages vs changes and repairs), totals, and built pages not published yet. Omit siteId for every workspace.

siteIdstring

Website id from status.

daysinteger

Days of history; default 30, at most 365.

Tasks, filtered. Compact rows; job_get has the detail.

siteIdstring

Website id from status.

statusplanned | queued | needs-input | researching | building | checking | ready | publishing | live | blocked | failed | cancelled | active

One status, or active (planning, building, checking, shipping).

formatarticle | alternatives | listicle | tool | marketing | programmatic | repair

Content type.

daystring

Scheduled day, YYYY-MM-DD in the Mac’s local time.

searchstring

Text in the title or keyword.

limitinteger

Rows to return; default 50, at most 500.

offsetinteger

Rows to skip.

One task in full: the facts an owner sees (keyword, volume, difficulty, what gets built), brief, plan, questions, checks, review, events, worktree path, branch and changed files. File contents are in the worktree; includeArtifacts returns stored contents for tasks without one.

idrequiredstring

Task id from jobs_list or calendar_get.

eventsinteger

Latest events to return; default 30.

includeArtifactsboolean

Return stored file contents.

fullboolean

Do not shorten long check and review output.

Audit work that can go on the calendar: keyword opportunities and findings with priority and whether a task already exists. sourceId feeds pipeline_enqueue.

siteIdrequiredstring

Website id from status.

unscheduledOnlyboolean

Hide work that already has a task; default true.

formatarticle | alternatives | listicle | tool | marketing | programmatic | repair

Content type.

limitinteger

Rows to return; default 50, at most 500.

offsetinteger

Rows to skip.

Put audit tasks on the content calendar. Validated work is scheduled automatically; use this for the rest. Existing bookings are kept.

siteIdrequiredstring

Website id from status.

sourceIdsrequiredstring[]

sourceId values from pipeline_candidates.

startDaterequiredstring

First day to fill, YYYY-MM-DD in the Mac’s local time; today or later.

batchboolean

Group up to 10 alternatives, comparison or marketing pages into one day.

Move a calendar task to a day, or pause it by omitting day. Moving a paused task resumes it.

idrequiredstring

Task id from jobs_list or calendar_get.

daystring

YYYY-MM-DD in the Mac’s local time; omit to pause.

Re-sync the calendar with saved audits and decisions.

No parameters.

Build all of a day’s tasks one after another with the execution model, including planning and review. The audit model stays unchanged. Nothing touches the website repository until pipeline_ship. A provider limit or a dropped connection pauses the run (today_get → pause); calling this again continues each task from its last finished step. Returns as soon as the work starts. Call wait, then status, to follow it.

siteIdrequiredstring

Website id from status.

daystring

YYYY-MM-DD in the Mac’s local time; omit for today.

executorobject

Model for all build steps, saved as the website’s execution model. Omit to keep the saved one.

executor.providerrequiredclaude | codex | opencode | ollama

Provider; check status for remaining quota.

executor.modelstring

Model id from settings_get; empty for the account default.

executor.effortstring

Reasoning effort; empty for the default.

Build today across selected workspaces, or all workspaces when siteIds is omitted. Runs one workspace after another, preserving each website’s day run. Empty and busy workspaces are skipped with reasons in activity_list. Use today_get without siteId for the combined view. Builds never publish. Returns as soon as the work starts. Call wait, then status, to follow it.

siteIdsstring[]

Website ids from status; omit for all workspaces.

daystring

YYYY-MM-DD in the Mac’s local time; omit for today.

executorobject

Execution model saved for each workspace that starts. Omit to keep each workspace’s saved execution model. Audit models stay unchanged.

executor.providerrequiredclaude | codex | opencode | ollama

Provider; check status for remaining quota.

executor.modelstring

Model id from settings_get; empty for the account default.

executor.effortstring

Reasoning effort; empty for the default.

Stop the day run after the task it is on.

siteIdrequiredstring

Website id from status.

Commit every built task to the website repository in build order, run the deploy command and verify each page live. Changes the live website: only when the owner asked to ship. Returns as soon as the work starts. Call wait, then status, to follow it.

siteIdrequiredstring

Website id from status.

Let AI plan eligible unbuilt tasks that sit on no day. Deferred, rejected and evidence-blocked ideas stay off the calendar until their audit decision or evidence changes: an agent with each website’s model puts each one back on the calendar from today, in value order, or removes it (cancelled, reason in skipped). Decisions are recorded in each task’s events. Progress is in status → triage. Only when the owner asked you to let AI plan the calendar. Returns as soon as the work starts. Call wait, then status, to follow it.

siteIdstring

Website id from status; omit for every workspace.

Create a build task from your own brief. It is not on the calendar; start it with job_run.

siteIdrequiredstring

Website id from status.

titlerequiredstring

Task title.

keywordrequiredstring

Target keyword; may be empty for repair.

formatrequiredarticle | alternatives | listicle | tool | marketing | programmatic | repair

Content type.

briefrequiredstring

What to build and why, with acceptance criteria.

opportunityIdstring

Workspace opportunity this task produces, from opportunities_list.

Edit a task that is not running or published. A changed brief resets it to planned.

idrequiredstring

Task id from jobs_list or calendar_get.

titlestring

New title.

briefstring

New brief.

scheduledAtstring

ISO date for tasks outside the calendar; empty clears. Calendar tasks move with pipeline_move.

Build one task now: the execution model plans, edits an isolated worktree and reviews, with website checks from the host. Returns as soon as the work starts. Call wait, then status, to follow it.

idrequiredstring

Task id from jobs_list or calendar_get.

Save answers to a task’s open questions (job_get → questions), even while another task builds. Once all are answered, resume it or queue it if the repository is busy. A queued response says whether it will join a pending workspace run or needs the next explicit Build today. Returns as soon as the work starts. Call wait, then status, to follow it.

idrequiredstring

Task id from jobs_list or calendar_get.

answersrequiredobject[]

One answer per open question key.

answers[].keyrequiredstring

Question key.

answers[].answerrequiredstring

The answer; pick one of the offered options when they fit.

Give a blocked or paused build concrete instructions, a missing answer or a correction, then retry the same task. Replies always save while other tasks build; a busy repository queues this retry, with queued:true and an explanation instead of claiming it started. Your comment remains in job_get. Never put credentials in feedback; configure them in the website environment. Returns as soon as the work starts. Call wait, then status, to follow it.

idrequiredstring

Task id from jobs_list or calendar_get.

feedbackrequiredstring

What the agent should do or the missing context, in plain words.

Stop a task that is planning, building, checking or shipping.

idrequiredstring

Task id from jobs_list or calendar_get.

Request changes to a page that is already live. Creates a new task that edits the same page from the owner’s feedback; start it with job_run and ship it like any build.

idrequiredstring

Task id from jobs_list or calendar_get.

feedbackrequiredstring

What should change on the live page, in plain words.

Re-run the website’s checks on a built task. On a task that was committed but not verified live, checks the live page again without rebuilding or deploying. Returns as soon as the work starts. Call wait, then status, to follow it.

idrequiredstring

Task id from jobs_list or calendar_get.

Commit and deploy one built task and verify it live. Changes the live website: only when the owner asked to publish. Prefer pipeline_ship for calendar tasks. Returns as soon as the work starts. Call wait, then status, to follow it.

idrequiredstring

Task id from jobs_list or calendar_get.

Open a built task’s preview window in the ClicksPal app for the owner to look at.

idrequiredstring

Task id from jobs_list or calendar_get.

Autopilot as its page shows it: settings, next run time, the task it will build next, what is running, why it is waiting (provider quota, other work), each workspace with tasks due, built pages waiting and whether its publishing schedule is due, the last runs, and the progress of AI answering questions (asks).

No parameters.

Change Autopilot: on/off, how often it builds one task (intervalHours), which workspaces take part (excludedSiteIds; new workspaces take part), when built pages publish (publish: immediate, batch after batchSize built changes or once the day is done, whichever comes first, dayDone when a workspace has built all of today’s tasks, manual), macOS notifications and whether an agent answers open questions first. Only the fields you send change. Autopilot runs one build at a time and only when nothing else is building. A stopped task is tried again automatically up to 3 times, at least an hour apart (timeouts as is, other stops after an AI decision), before it waits in Needs you. Read it with autopilot_get.

settingsrequiredobject

Autopilot settings.

settings.enabledboolean

true turns Autopilot on; the first task starts within a minute when nothing else is building.

settings.intervalHoursnumber

Hours between tasks: 0.25, 0.5, 1, 2, 4, 6 (0.25 = 15 minutes, 0.5 = 30 minutes).

settings.excludedSiteIdsstring[]

Website ids from status that do not take part. Replaces the saved list.

settings.publishimmediate | batch | dayDone | manual

When built pages go live. A workspace whose publication policy is review never publishes automatically.

settings.batchSizeinteger

Built changes a workspace collects before batch publishing, 1–50.

settings.notifyobject

macOS notifications.

settings.notify.builtboolean

A page or tool was built.

settings.notify.publishedboolean

Pages were published, or publishing failed.

settings.notify.needsYouboolean

A task needs an answer or stopped.

settings.answerQuestionsboolean

true lets an agent answer or dismiss open questions (as asks_resolve) before each task.

Have Autopilot build its next task now (autopilot_get → next), then publish what its schedule says is due. Refused while anything else is building. Returns as soon as the work starts. Call wait, then status, to follow it.

No parameters.

Let AI handle the questions waiting on the owner: an agent with each website’s execution model answers every open task and setup question (picking an offered option), or dismisses a task it cannot build truthfully. Answered tasks are queued for today; dismissed tasks are cancelled with the reason in skipped. Decisions are recorded in each task’s events. Progress is in autopilot_get → asks. Only when the owner asked you to let AI answer. Returns as soon as the work starts. Call wait, then status, to follow it.

siteIdstring

Website id from status; omit for every workspace.

Search Console clicks, impressions, CTR and average position from the latest saved report, grouped by query, page or date.

siteIdrequiredstring

Website id from status.

groupByquery | page | date

Grouping; default query.

searchstring

Text in the query or page.

limitinteger

Rows to return; default 50, at most 500.

offsetinteger

Rows to skip.

AI visibility: the measured buyer-question panel, and what ChatGPT and Gemini answered — whether the product was mentioned or cited, and which competitors were.

siteIdrequiredstring

Website id from status.

includeAnswersboolean

Return full answer text.

limitinteger

Rows to return; default 50, at most 500.

offsetinteger

Rows to skip.

Generate and measure a keyword panel (search) or buyer-question panel in ChatGPT and Gemini (ai). Spends the saved DataForSEO budget; cached measurements are reused. Returns as soon as the work starts. Call wait, then status, to follow it.

siteIdrequiredstring

Website id from status.

channelrequiredsearch | ai

Which panel.

Ask ChatGPT and Gemini the saved buyer questions and record mentions and citations. Spends the saved DataForSEO budget; cached measurements are reused. Returns as soon as the work starts. Call wait, then status, to follow it.

siteIdrequiredstring

Website id from status.

Record an AI answer you observed yourself as manual evidence.

siteIdrequiredstring

Website id from status.

promptrequiredstring

The question asked.

answerrequiredstring

The full answer text.

citationsrequiredstring[]

URLs the answer cited.

competitorsrequiredstring[]

Competitor names the answer mentioned.

Pull the latest finalized 28 days of query and page performance from Google Search Console. Returns as soon as the work starts. Call wait, then status, to follow it.

siteIdrequiredstring

Website id from status.

Import a Search Console performance CSV.

siteIdrequiredstring

Website id from status.

csvrequiredstring

CSV text with query, page, clicks, impressions, ctr and position columns.

startDaterequiredstring

Report start, YYYY-MM-DD.

endDaterequiredstring

Report end, YYYY-MM-DD.

Paid research spend: totals per month and endpoint against the saved budgets, and the latest ledger rows.

siteIdstring

Only this website.

limitinteger

Rows to return; default 50, at most 500.

offsetinteger

Rows to skip.

ClicksPal AI calls and reported tokens for today and the last seven days, plus actual provider quota windows. Unknown token usage remains unknown.

No parameters.

App defaults, research budgets, which accounts are connected (never the credentials), and every AI provider with its models, efforts and remaining quota.

No parameters.

Change app-wide defaults and research budgets. Only the fields you send change.

settingsrequiredobject

Settings.

settings.providerclaude | codex | opencode | ollama

Default provider.

settings.modelstring

Default model.

settings.effortstring

Default effort.

settings.maxConcurrentJobsinteger

1–4.

settings.reserveQuotaPercentnumber

Provider quota to leave unused, 0–90.

settings.monthlyBudgetUsdnumber

Monthly DataForSEO budget.

settings.keywordRequestLimitUsdnumber

Cap per keyword request.

settings.visibilityRequestLimitUsdnumber

Cap per AI-answer request.

settings.ollamaUrlstring

Loopback Ollama URL.

settings.providerChoicesobject

Per-provider {model, effort}.

settings.secondaryAgentobject

Explicit backup for capacity limits.

settings.secondaryAgent.providerrequiredclaude | codex | opencode | ollama

Provider; check status for remaining quota.

settings.secondaryAgent.modelstring

Model id from settings_get; empty for the account default.

settings.secondaryAgent.effortstring

Reasoning effort; empty for the default.

settings.disableSecondaryAgentboolean

Clear the backup provider.

settings.setupCompleteboolean

First-run checklist was completed.

Store DataForSEO or Google OAuth client credentials in the Mac’s encrypted storage. Write-only: saved values are never returned. An empty string deletes one.

dataforseoLoginstring

DataForSEO login.

dataforseoPasswordstring

DataForSEO password.

gscClientIdstring

Google OAuth client id.

gscClientSecretstring

Google OAuth client secret.

Start Google sign-in for Search Console in the owner’s browser. The owner finishes it; settings_get shows the result. Returns as soon as the work starts. Call wait, then status, to follow it.

No parameters.

Connect Search Console with a Google service-account key file. The key is stored encrypted and checked against the API first. Returns as soon as the work starts. Call wait, then status, to follow it.

pathrequiredstring

Absolute path to the service-account .json key.

Re-detect installed AI providers, their models and remaining quota. A newly connected provider is asked a tiny test question; settings_get → providers[].test shows the result.

retestOldboolean

Also re-test providers whose last passed test is over 30 days old.

Ask one connected provider a tiny test question with a model and reasoning level. Only a passed test (providers[].test.state ok in settings_get) makes it Ready. OpenCode with no model tries one model per account (Go, other accounts, then the Console's free model, never a paid one) and saves the first that answers as its model (providers[].test.model). Returns at once; the test takes up to two minutes. Ollama has none.

providerrequiredclaude | codex | opencode | ollama

Provider.

modelstring

Model id; omit for the saved choice, empty for the provider default.

effortstring

Reasoning level; omit for the saved choice.