> For the complete documentation index, see [llms.txt](https://help.usejimo.com/docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://help.usejimo.com/docs/integrations/mcp/tools-reference.md).
# Tools reference
Everything the assistant can do, grouped by the permission that unlocks it.
Tool names are what the assistant sees. You rarely need them, but they appear in error messages and in the assistant's own explanations, so they are listed here for reference.
## Read Project Data and Analytics
These tools are read-only: they never change anything, never ask for confirmation, and do not count against the monthly allowance. Test questions are free and limited to about 30 per project per hour.
### Experiences, Users, and Feedback
| Tool | What it does |
| ---------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `intel_get_project_overview` | The project at a glance: users and activity, what is live by type, drafts, experiences expiring soon, best performers, running experiments, activation summary. |
| `intel_list_experiences` | Inventory of tours, surveys, banners, hints, checklists, resource centers, and posts with their delivery status. Can be filtered by tag. |
| `intel_get_experience_performance` | One experience's funnel: completion and engagement rates, with ordered step-by-step drop-off. |
| `intel_get_activation_funnel` | Step-by-step activation for a success tracker, optionally for one segment. |
| `intel_compare_cohort_adoption` | How up to five segments adopt one experience, against the all-users baseline. |
| `intel_get_experiment_results` | A/B results per variant, lift over control, and whether the difference is significant yet. |
| `intel_list_issues` | Where experiences break for real users: target element missing, hidden, or ambiguous. |
| `intel_get_user_adoption_profile` | One end user: identity, segments, every experience seen with status, activation progress. |
| `segment_list` | The project's segments with approximate sizes. |
| `feedback_list_responses` | Survey and NPS responses, newest first. |
| `content_get_experience` | Reads one experience back as stored: copy step by step, the editable structure (steps, triggers, survey logic, checklist tasks, resource center groups), audience, display rules, theme, tags, a dashboard link, and a signed preview link that opens the draft in your app. Experiences built in the dashboard can be richer than this view shows, and the tool says so when parts are not represented. |
| `content_list_templates` | The dashboard's template gallery, with the steps that need an element of your app. |
| `content_list_themes` | The project's themes, and which one new experiences start in. |
| `content_list_tags` | The experience tags and the changelog post tags, with how many experiences carry each. |
### Jimo Agent
| Tool | What it does |
| --------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `agent_get_overview` | Where the agent stands, and the setup gaps still open. The place to start. |
| `agent_list_pages` | Every screen the agent knows. Pass a URL to see which pages would load there. |
| `agent_get_page` | One page: description, rules, elements by name and purpose, suggestions, skills, and the screenshot link. When the assistant cannot use a browser to capture an element, experience steps can be pinned to these elements. |
| `agent_test_url_rules` | Tries rules against real URLs before anything is saved. |
| `agent_list_skills` | Every skill, with its type, page, uses, and whether it needs review (10 or more decided runs and under 70% resolved). |
| `agent_get_skill` | One skill: instructions, the values it asks for, and the suggestions that start it. |
| `agent_get_skill_performance` | Why a skill fails, split into structural causes (a missing element, an unscanned page), which rewriting the instructions does not fix, and instruction issues, which it may. |
| `agent_list_knowledge_sources` | Everything the agent is trained on, with training status. |
| `agent_get_knowledge_source` | Training progress, the pages a website actually read, and the ones it could not. |
| `agent_test_question` | Asks the real agent. Returns the answer, what it looked up (it may not have used all of it), the skill it would offer, the hand-off to a human it would start, and the first page action it would take. That action is not run. Pass a page URL, and the text visible there, to ask from a given screen. |
| `agent_get_performance` | Usage, success rate, and escalations by reason and topic, against the previous period. Success rate is likes against dislikes: conversations nobody reacted to do not count. |
| `agent_get_conversation_insights` | What end users asked the Jimo Agent, plus a needs-attention feed: unanswered questions, disliked answers, abandoned skill runs. |
| `agent_list_conversations` | What end users asked, filtered by text, reaction, escalation, or page. |
| `agent_get_conversation` | One transcript. |
## Manage Experiences
| Tool | What it does | Asks to confirm? |
| ------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | --------------------------------------------------------------------- |
| `content_draft_experience` | Creates a draft of any kind: changelog post, banner, modal, tour, survey, hint, checklist, or resource center, from text, from steps, or from a gallery template. Steps pinned to an element take an element captured in your app with `anchor:capture`, or, when the assistant cannot use a browser, an element from a screen the agent has scanned. Tours can carry progress triggers, checklist tasks their buttons and completion conditions, and any button can launch another experience. On a button or a link, the draft review turns a tooltip into a cursor. Drafts are invisible to end users and start in the project's default theme. Images are attached by public URL. | No (draft only) |
| `content_update_experience` | Edits any experience, draft or published, step by step: copy, buttons, steps (insert, move, remove), survey logic, checklist tasks, resource center groups and actions, tour progress triggers, and the element a step is pinned to. Up to 50 edits per call, applied all together or not at all. Everything it does not touch stays as it was. On a published experience, the edit is refused if the experience changed since the assistant read it, and removing a survey step hides it. | On a published experience, or a draft also worked on in the dashboard |
| `content_set_display_rules` | Sets where and when an experience shows: pages, domains, trigger (page load with a delay, exit intent, tracked event, element click, **On shared URL**, **On manual trigger**, or **No automatic trigger**), recurrence, priority, and devices. For an experience another one launches, use **No automatic trigger**. | Yes |
| `content_set_appearance` | Applies one of the project's themes to an experience, and sets where each step appears. The theme replaces styling made in the builder. | On a published experience, or a draft also worked on in the dashboard |
| `content_set_experience_tags` | Adds or removes tags on an experience, and can create a missing tag. One unit per call, however many tags. | Only on a published changelog post |
| `content_manage_tags` | Creates, renames, recolors, or deletes a tag. | When renaming or deleting |
| `content_manage_experiment` | Runs an A/B test: creates variant B as a copy you then edit, sets the traffic split, starts the test once the experience is live, and ends it with a winner or cancels it. Ending shows the current results and warns when they are not significant yet. Creating and ending a test needs a signed-in Jimo user: an API key cannot. Not for changelog posts. | Yes |
| `content_publish_experience` | Makes a draft visible in chosen environments, or re-scopes a published experience's environments. Posts skip environment scoping, they are portal content. A publish your plan does not allow is refused at once. | Yes |
| `content_set_experience_state` | Pauses or resumes a published experience. Not for posts. | Yes |
| `content_set_audience` | Sets who an experience targets (saved segments or all users) and when it runs (start and expiry). | Yes |
| `content_translate_experience` | Translates every step and block of an experience with Jimo's smart translate, formatting kept. Works on every experience type, draft or published. Adds missing languages by default, or regenerates existing ones after the base copy changed. | Yes, except on a draft the assistant created and nobody else touched |
| `content_set_translation` | Writes one language's title, body, or button label by hand, to fix a wording or supply a translation. Posts, banners, and modals only. | Yes, except on a draft the assistant created and nobody else touched |
| `content_delete_translation` | Removes one language, so end users with that language see the base copy. | Yes |
| `content_delete_experience` | Deletes a draft, or a published experience or post, together with what it collected (survey responses, comments, reactions). It does not archive and cannot be undone. Deleting a published item needs the delete permission. The preview offers pausing instead when that is enough. | Yes |
One request can wire a whole journey: task buttons, progress triggers and survey logic can each launch another experience.
## Teach My Jimo Agent
| Tool | What it does | Asks to confirm? |
| -------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | ---------------- |
| `agent_save_custom_answer` | Saves a curated question and answer pair. Live for end users within minutes. Saving the same question again updates its answer. | Yes |
| `agent_add_knowledge_source` | Imports up to 10 documents per call from text the assistant already holds. Re-importing with the same document id replaces the old version. | Yes |
| `agent_add_website_source` | Trains on a help center or docs site from its sitemap (same host, under that address, without following links), or on an exact list of pages. The preview shows the pages found, up to 500, and the existing website sources that already read some of them, before anything trains. | Yes |
| `agent_add_zendesk_source` | Trains on the connected Zendesk help center, all published articles (up to 500) or a selection. If Zendesk is not connected, it returns the dashboard link where an admin connects it. | Yes |
| `agent_refresh_knowledge_source` | Re-crawls an existing URL source now instead of waiting for the schedule. | No |
| `agent_delete_knowledge_source` | Removes a custom answer, an imported document, a website, a Zendesk help center, or an uploaded file. A source that is still training cannot be deleted. | Yes |
## Build and Run My Jimo Agent
| Tool | What it does | Asks to confirm? |
| ----------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------ |
| `agent_update_settings` | Instructions, language, attributes the agent may see, launcher targeting, devices, segments, accent and theme, escalation, and how often websites are re-read. Left-out settings stay as they are. Does not switch the launcher on or off. | Yes |
| `agent_set_launcher_state` | Puts the launcher live or pauses it. The preview lists what is still missing. | Yes, unless it is already in that state |
| `agent_create_read_all_pages` | Up to 10 Read all screens from a name, a description in users' words, and URL rules. No snippet needed to create them, they take effect once the Jimo snippet runs on the site. Each one comes back with a screenshot link. If one page is refused, the whole batch is refused with the reason for each. | Yes |
| `agent_update_page` | Name, description, rules, environments. Also the step after a scan, to cover every record of the screen. | Yes |
| `agent_delete_page` | Deletes a page and every skill on it. Needs **Delete pages and skills**. | Yes |
| `agent_set_page_suggestions` | The ordered suggested questions on a page, at most four active. A suggestion can start a skill. Suggestions left out of the call are deleted. | Yes |
| `agent_start_page_scan` | A dashboard link that opens the scanner. Writes nothing itself: saving the scan in the dashboard adds the page, replaces its scan, or converts it to the other mode, and says which first. | No |
| `agent_create_skills` | Up to 10 skills, always created inactive. If one skill is refused, the whole batch is refused with the reason for each. | No |
| `agent_update_skill` | Name, description, instructions, type, or move to another page. Live users get the new version within about 30 seconds. | Yes when the skill is live. An inactive skill is updated directly. |
| `agent_set_skill_state` | Activates or deactivates up to 20 skills. | Yes, unless nothing would change |
| `agent_delete_skill` | Deletes a skill. Deactivating keeps it. Needs **Delete pages and skills**. | Yes |
{% hint style="info" %}
Deleting a page or a skill also turns off the suggestions that started it, so the launcher stops offering a skill that is gone.
{% endhint %}
## Next Steps
* [What to Ask](/docs/integrations/mcp/what-to-ask.md): you never call these tools yourself, just ask in plain language.
* [Permissions & Safety](/docs/integrations/mcp/permissions-and-safety.md): how the confirmation step works.
