> ## Documentation Index > Fetch the complete documentation index at: https://agentheya.mintlify.site/llms.txt > Use this file to discover all available pages before exploring further. # feed_monitors > Recurring RSS and X (Twitter) watchers. Each monitor polls a source and can react to new items (off / all / classified), optionally delivering results to… # `feed_monitors` — Feed monitors Recurring RSS and X (Twitter) watchers. Each monitor polls a source and can react to new items (off / all / classified), optionally delivering results to email or the inbox. **Section key** `feed_monitors` · **Shape** array · **Tier** All plans\ **Edit in the dashboard:** [Feeds page](/sidebar-menu/feeds)\ **Related:** [Schedules](/sidebar-menu/schedules) > List of feed monitor objects. ## Fields (per item) | Field | Required | Type | Allowed values / constraints | Default | Description | | ----------------------------- | -------- | --------------- | --------------------------------- | ------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | `id` | **yes** | string | — | — | Stable id for this monitor (10+ char random). Used as the dedup/cursor key and to match the feed\_monitor DB row. | | `enabled` | no | boolean | — | — | Whether the poller should fire this monitor. | | `kind` | **yes** | enum | one of `rss`, `twitter` | — | Source type: 'rss' (feed\_url) or 'twitter' (handle). | | `feed_url` | no | string | — | — | For kind='rss': the full http(s) URL of the RSS/Atom feed. | | `handle` | no | string | — | — | For kind='twitter': the username to watch (with or without a leading '@'). | | `query` | no | string | — | — | For kind='twitter': optional keyword/phrase to further filter the user's tweets. | | `poll_interval_minutes` | no | integer | — | — | How often to poll, in minutes. Clamped to the platform-configured minimum. | | `react_mode` | no | enum | one of `off`, `all`, `classified` | — | How to react to new items. 'off' = log only; 'all' = run the agent on every new item; 'classified' = run an LLM classifier first and react only when it returns action='react'. | | `classifier_prompt` | no | string | — | — | For react\_mode='classified': owner guidance describing which items are relevant enough to react to. | | `classifier_action_allowlist` | no | array of string | — | — | For react\_mode='classified': restrict the actions the classifier may pick. Subset of \["react", "drop"]. | | `react_instructions` | no | string | — | — | The task the agent runs when it reacts to a new item. The item's title/author/link/content are appended automatically. | | `delivery` | no | object | — | — | Where the agent's reaction output goes. 'inbox' is implicit (the reaction always mirrors into the owner Inbox); 'email' also sends the output to a target address. Item fields below. | | `label` | no | string | — | — | Optional human-readable label for the dashboard. | | `preProcessor` | no | object | — | — | Code-first pre-processor: owner JS/Python that runs per new item BEFORE the LLM classifier and returns a verdict. A 'drop' or 'store' here skips the classifier; 'pass\_to\_agent' falls through to react\_mode. See \$lib/server/event-pipeline. Item fields below. | | `interpreters` | no | array of object | — | — | Payload interpreters: each is tested in order against a new item and the first match injects its hint into the agent's context. Part of the unified incoming-event pipeline. See \$lib/shared/event-pipeline. Item fields below. | | `notes` | no | string | — | — | Private operator notes about this monitor. Never sent to the model. | | `collapsed` | no | boolean | — | — | UI hint — start this entry collapsed in the dashboard editor. | ### `delivery` fields Where the agent's reaction output goes. 'inbox' is implicit (the reaction always mirrors into the owner Inbox); 'email' also sends the output to a target address. | Field | Required | Type | Allowed values / constraints | Default | Description | | -------- | -------- | ------ | ------------------------------- | ------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | `kind` | no | enum | one of `none`, `email`, `inbox` | — | Controls delivery of the agent's reaction output. 'none' disables delivery; 'email' sends output to the target address; 'inbox' (default) mirrors reactions into the owner's Inbox. | | `target` | no | string | — | — | Email address when kind='email'. | ### `preProcessor` fields Code-first pre-processor: owner JS/Python that runs per new item BEFORE the LLM classifier and returns a verdict. A 'drop' or 'store' here skips the classifier; 'pass\_to\_agent' falls through to react\_mode. See \$lib/server/event-pipeline. | Field | Required | Type | Allowed values / constraints | Default | Description | | ---------- | -------- | ------- | ----------------------------- | ------- | ------------------------------------------------------------------ | | `enabled` | no | boolean | — | — | Whether the pre-processor runs. | | `language` | no | enum | one of `python`, `javascript` | — | Code language for the handle(event, ctx) function. | | `code` | no | string | — | — | Source defining handle(event, ctx) which returns a verdict object. | ### `interpreters` items Payload interpreters: each is tested in order against a new item and the first match injects its hint into the agent's context. Part of the unified incoming-event pipeline. See \$lib/shared/event-pipeline. | Field | Required | Type | Allowed values / constraints | Default | Description | | ----------------- | -------- | ------ | ----------------------------- | ------- | -------------------------------------------------------------------------------------------- | | `name` | no | string | — | — | Human-readable label for this interpreter. | | `matchType` | no | enum | one of `json_fields`, `regex` | — | How this interpreter matches: presence of JSON dot-paths, or a regex over the raw item text. | | `matchFields` | no | string | — | — | For matchType='json\_fields': newline/comma-separated dot-paths that must all be present. | | `matchRegex` | no | string | — | — | For matchType='regex': pattern tested against the serialized item. | | `matchRegexFlags` | no | string | — | — | Optional regex flags, e.g. 'i'. | | `hint` | no | string | — | — | Instruction prepended to the agent's context when this interpreter matches. |