This is the full developer documentation for Engagent # Engagent documentation > Engagent finds the LinkedIn posts that are worth a comment from you each day. Use it in the app, or from your AI agent. ## What Engagent does [Section titled “What Engagent does”](#what-engagent-does) Each day, Engagent searches LinkedIn for posts that match your keywords. It ranks the posts for you and tells you why each one ranks. You write the comment and you post it. Engagent does not post to LinkedIn for you. Engagent also shows who engaged with a post, and scores those people against your ideal customer. You can send the best of them to your outreach tool. ## Two ways to use it [Section titled “Two ways to use it”](#two-ways-to-use-it) In the app Open the app each morning. Work down the list of posts, write your comments, and mark each post as posted. [Open the app](https://app.myengagent.com) From your AI agent Connect Claude, ChatGPT or another AI agent. The agent gets the posts for the day and drafts each comment in your voice. [See the tools an agent can use](/reference/mcp-tools/all-tools/) ## For developers [Section titled “For developers”](#for-developers) [API reference](https://api.myengagent.com)Every API route, made from the code. [MCP tools](/reference/mcp-tools/all-tools/)Each tool an AI agent can call, with its inputs. ## Built for agents [Section titled “Built for agents”](#built-for-agents) Every page here is also plain Markdown. Add `.md` to the address of a page, or use **Copy page** at the top of the page. * [`/llms.txt`](/llms.txt) lists what these docs hold. * [`/llms-full.txt`](/llms-full.txt) is every page in one file. # All MCP tools > The 9 tools an AI agent has in Engagent, and what each one does. This page is made from the code, so it matches the server your AI agent connects to. An MCP tool is one thing your AI agent can ask Engagent to do. You do not call a tool yourself. You ask your AI agent in plain words, and it picks the tool. ## Each day [Section titled “Each day”](#each-day) | Tool | Does | Changes data | | -------------------------------------------------------------------------------- | ----------------------------------------------- | ------------ | | [`engagent_get_status`](/reference/mcp-tools/engagent-get-status/) | Workspace status | No | | [`engagent_list_posts`](/reference/mcp-tools/engagent-list-posts/) | Posts to comment on | No | | [`engagent_get_post`](/reference/mcp-tools/engagent-get-post/) | One post in full | No | | [`engagent_get_comment_style`](/reference/mcp-tools/engagent-get-comment-style/) | How the user comments | No | | [`engagent_update_post`](/reference/mcp-tools/engagent-update-post/) | Mark a post posted, skipped, snoozed or pending | Yes | | [`engagent_exclude_author`](/reference/mcp-tools/engagent-exclude-author/) | Never show this author again | Yes | ## Set up a workspace [Section titled “Set up a workspace”](#set-up-a-workspace) | Tool | Does | Changes data | | ---------------------------------------------------------------------------- | --------------------------------------- | ------------ | | [`engagent_propose_profile`](/reference/mcp-tools/engagent-propose-profile/) | Propose the profile from a website | No | | [`engagent_save_profile`](/reference/mcp-tools/engagent-save-profile/) | Save the profile and keywords | Yes | | [`engagent_finish_setup`](/reference/mcp-tools/engagent-finish-setup/) | Finish setup and start the first search | Yes | ## Prompts [Section titled “Prompts”](#prompts) A prompt is a set of instructions that Engagent gives your AI agent for a routine. | Prompt | Does | | ------------------------------------------------------------------- | ------------------------- | | [`daily-engagement`](/reference/mcp-tools/prompt-daily-engagement/) | Today’s LinkedIn comments | # engagent_exclude_author > Never show this author again. | | | | -------------- | --------------------------------------------------------------- | | Changes data | Yes | | Who can use it | A workspace admin or editor, in a workspace with an active plan | ## What the AI agent is told [Section titled “What the AI agent is told”](#what-the-ai-agent-is-told) This is the description your AI agent reads, word for word. > Stop showing posts by this post’s author, from the next search on. Confirm with the user first; they can undo it in Settings > Excluded profiles. ## Arguments [Section titled “Arguments”](#arguments) | Argument | Type | Needed | What | | -------- | ---- | ------ | ------------------------------------- | | `id` | id | yes | The post id from engagent_list_posts. | # engagent_finish_setup > Finish setup and start the first search. | | | | -------------- | ----------------- | | Changes data | Yes | | Who can use it | A workspace admin | ## What the AI agent is told [Section titled “What the AI agent is told”](#what-the-ai-agent-is-told) This is the description your AI agent reads, word for word. > Finish setup: schedule the nightly search and start the first one now. Call it once, after engagent_save_profile has saved the profile and keywords. If the plan is unpaid the nightly search waits for payment, but the first one still runs for a preview. ## Arguments [Section titled “Arguments”](#arguments) No arguments. # engagent_get_comment_style > How the user comments. | | | | -------------- | --------------------------------------------------------------- | | Changes data | No. It only reads | | Who can use it | Any member of the workspace, in a workspace with an active plan | ## What the AI agent is told [Section titled “What the AI agent is told”](#what-the-ai-agent-is-told) This is the description your AI agent reads, word for word. > What you need to draft a comment as the user: their tone, their instructions, the business, the instruction for one post’s topic and level (pass the post id), and comments they posted recently as examples. Call it once before you draft. ## Arguments [Section titled “Arguments”](#arguments) | Argument | Type | Needed | What | | --------- | ---- | ------ | ----------------------------------------------------------------------------- | | `post_id` | id | no | Optional post id, to include the instruction for that post’s topic and level. | # engagent_get_post > One post in full. | | | | -------------- | --------------------------------------------------------------- | | Changes data | No. It only reads | | Who can use it | Any member of the workspace, in a workspace with an active plan | ## What the AI agent is told [Section titled “What the AI agent is told”](#what-the-ai-agent-is-told) This is the description your AI agent reads, word for word. > The full text of a post from engagent_list_posts, with reactions, comments and shares, its tags and the ICP reason. Read this before drafting a comment. ## Arguments [Section titled “Arguments”](#arguments) | Argument | Type | Needed | What | | -------- | ---- | ------ | ------------------------------------- | | `id` | id | yes | The post id from engagent_list_posts. | # engagent_get_status > Workspace status. | | | | -------------- | --------------------------- | | Changes data | No. It only reads | | Who can use it | Any member of the workspace | ## What the AI agent is told [Section titled “What the AI agent is told”](#what-the-ai-agent-is-told) This is the description your AI agent reads, word for word. > The workspace at a glance: whether the nightly search ran or is running, posts waiting, comments posted this week, streak, plan, the saved search keywords. While setup is unfinished it also lists the setup steps, what is missing and the next one, with the browser URL for anything that needs it (payment). ## Arguments [Section titled “Arguments”](#arguments) No arguments. # engagent_list_posts > Posts to comment on. | | | | -------------- | --------------------------------------------------------------- | | Changes data | No. It only reads | | Who can use it | Any member of the workspace, in a workspace with an active plan | ## What the AI agent is told [Section titled “What the AI agent is told”](#what-the-ai-agent-is-told) This is the description your AI agent reads, word for word. > Today’s LinkedIn posts worth commenting on, best first (most expected ICP matches among the engagers, then newest). Each has an id, author, a 300-character snippet, the post URL and why it ranks. Use engagent_get_post for the full text. Says when the nightly search is still running. ## Arguments [Section titled “Arguments”](#arguments) | Argument | Type | Needed | What | | ------------ | ------ | ------ | ---------------------------------------------------- | | `since_days` | number | no | Posts queued in the last N days, 1 to 30. Default 7. | | `status` | text | no | pending (default), posted, skipped, snoozed or all. | | `limit` | number | no | How many to return, 1 to 25. Default 10. | | `offset` | number | no | Items to skip, for the next page. Default 0. | # engagent_propose_profile > Propose the profile from a website. | | | | -------------- | ----------------- | | Changes data | No. It only reads | | Who can use it | A workspace admin | ## What the AI agent is told [Section titled “What the AI agent is told”](#what-the-ai-agent-is-told) This is the description your AI agent reads, word for word. > Read the user’s website and propose the whole setup: industry, product description, tone, keywords and ideal customer profile. Nothing is saved. Show it to the user, adjust, then engagent_save_profile. ## Arguments [Section titled “Arguments”](#arguments) | Argument | Type | Needed | What | | -------------------- | ---- | ------ | --------------------------------------------------------- | | `website_url` | text | yes | The website, for example acme.com. | | `company_name` | text | no | The company name, when the site does not make it obvious. | | `additional_context` | text | no | Anything the user told you that the site does not say. | # engagent_save_profile > Save the profile and keywords. | | | | -------------- | ----------------- | | Changes data | Yes | | Who can use it | A workspace admin | ## What the AI agent is told [Section titled “What the AI agent is told”](#what-the-ai-agent-is-told) This is the description your AI agent reads, word for word. > Save the business profile, the ideal customer profile and the search keywords, in setup or later. Only the fields you pass change; the rest keep their saved values. A keywords list replaces the saved one, so pass every keyword to keep. ## Arguments [Section titled “Arguments”](#arguments) | Argument | Type | Needed | What | | ---------------------- | ---- | ------ | ------------------------------------------------------------------------------------------- | | `industry` | text | no | The industry, a few words. | | `product_description` | text | no | What the business does, 20 to 50 words, starting ‘We help…’. | | `niche_description` | text | no | The niche or audience, free text. | | `tone` | text | no | professional, conversational, bold or analytical. | | `tone_notes` | text | no | Notes on the tone, free text. | | `comment_instructions` | text | no | Standing instructions for every comment, free text. | | `own_linkedin_url` | text | no | The user’s own LinkedIn profile URL. | | `icp_industries` | list | no | ICP industries in plain words; each is matched to one of Engagent’s 20 industry groups. | | `icp_company_sizes` | list | no | ICP company sizes, from: 1-10, 11-50, 51-200, 201-1,000, 1,001-5,000, 5,001+. | | `icp_locations` | list | no | ICP locations (countries or regions). | | `icp_roles` | list | no | ICP roles or job titles. | | `keywords` | list | no | LinkedIn search keywords for the nightly search: the full list, 1 to 10, each 1 to 3 words. | # engagent_update_post > Mark a post posted, skipped, snoozed or pending. | | | | -------------- | --------------------------------------------------------------- | | Changes data | Yes | | Who can use it | A workspace admin or editor, in a workspace with an active plan | ## What the AI agent is told [Section titled “What the AI agent is told”](#what-the-ai-agent-is-told) This is the description your AI agent reads, word for word. > Set what happened to a post. posted: the user commented on LinkedIn; pass the exact text they posted when you know it, it counts towards the streak. skipped: they chose not to comment. snoozed: hide it until a UTC time. pending: put a posted, skipped or snoozed post back in the list. ## Arguments [Section titled “Arguments”](#arguments) | Argument | Type | Needed | What | | -------- | ------------- | ------ | ------------------------------------------------------------------------------------- | | `id` | id | yes | The post id from engagent_list_posts. | | `status` | text | yes | posted, skipped, snoozed or pending. | | `text` | text | no | With posted: the comment as the user posted it. Omit when unknown. | | `until` | date and time | no | With snoozed: when to bring it back, UTC, ISO 8601, for example 2026-10-22T08:00:00Z. | # daily-engagement > The morning routine. The morning routine: fetch today’s posts from Engagent, pick the best, draft a comment for each in the user’s voice, then record what they posted. ## Arguments [Section titled “Arguments”](#arguments) | Argument | Type | Needed | What | | -------- | ------ | ------ | ------------------------------------------ | | `count` | number | no | How many posts to work through. Default 5. | ## What the AI agent is told [Section titled “What the AI agent is told”](#what-the-ai-agent-is-told) This is the text of the prompt, with every argument left at its default. ```text Run my Engagent routine for today. 1. Call engagent_list_posts. If it says the nightly search is still running, tell me and stop. Otherwise pick the 5 best posts to comment on. For each, one line: who, what the post says, and why it is worth a comment (the ICP reason). 2. Call engagent_get_comment_style once. For each pick, call engagent_get_post, then draft one comment in my voice: first person, two to four sentences, one idea, no hashtags, no links, no pitch, no mention of AI or Engagent. Add something the post did not say. 3. Show me the drafts with the post links. I will post the ones I like on LinkedIn myself. 4. When I tell you which I posted (and any edits I made), call engagent_update_post for each with status posted and the exact text I used, and with status skipped for the ones I passed on. ```