briefpanel
Server Details
Monitor web pages and read item-level updates: the field that moved, with its old and new value.
- Status
- Healthy
- Uptime
- 100.0% over the last 24 hours
- OAuth
- Works in Glama
- Last Tested
- Transport
- Streamable HTTP · MCP 2024-11-05
- URL
- ChatGPT
- –
- Claude
- –
TDQS
Scored across 14 tools
Source management, entry retrieval, credit balance, and identity tools have distinct purposes. The main ambiguity is between get_entry_changes and get_entry_updates, whose history-oriented descriptions do not clearly explain when one should be preferred; get_source_changes is easier to distinguish by its account-or-source scope.
Nearly all tools follow a consistent snake_case verb_noun pattern with predictable source and entry resource names. The standalone whoami is a minor deviation from that convention.
Fourteen tools is reasonable for monitoring sources, reading entries and histories, managing monitoring state, and checking account credits. The surface is substantial without becoming broadly unfocused.
The set covers source creation, inspection, settings updates, pause/resume, extraction relearning, entry reads, change history, and account status. The main lifecycle gap is the absence of a permanent delete or unmonitor operation, although pause_source provides a workaround for stopping checks.
Publisher details
- Operator
- Briefpanel · Publisher source
- Operator website
- https://briefpanel.com · Publisher source
- Vendor relationship
- First-party · Publisher source
- Documentation
- https://briefpanel.com/mcp · Publisher source
- Privacy policy
- Not provided
- Trust center
- Not available
- Restrictions
- Requires a Briefpanel account. Usage spends prepaid pay-as-you-go credits; no subscription required. · Publisher source
Available Tools
14 toolsadd_sourceMonitor a new pageAIdempotentInspect
Start monitoring a web page for the user. One page per URL: if the account already monitors this URL — under any instructions — no second page is started. The settings this call carries are applied to the page that already exists instead, and the response says which case you got: created is true for a new page, false for one that was already monitored, where updatedSettings lists what this call changed on it. Each check spends one credit per page it reads — the source page, plus each item page it follows.
Inputs: url (required), userPrompt, monitoringFrequency, notificationsEnabled.
get_credit_balanceGet credit balanceARead-onlyIdempotentInspect
Get the user's remaining BriefPanel credits. A check spends one credit per page it reads — the source page, plus each item page it follows.
Inputs: no named parameters.
get_entryGet an entry in fullARead-onlyIdempotentInspect
Read one item in full — everything list_source_entries returns for it plus the prose the list omits: summary (what the page says about the item — the article's text, the quote, the job posting) and author. For every non-priced kind (quote, news article, blog post, job posting) these two fields ARE the entry's content, and list_source_entries reduces it to a title. Also carries language, the locale the prose was written in. Use it after list_source_entries names an entry worth reading.
Inputs: entryId (required).
get_entry_changesGet entry changesARead-onlyIdempotentInspect
Get one entry's change timeline grouped by fetch, including structured evidence availability. A page can be empty or shorter than limit while more results remain; continue with pagination.nextCursor until pagination.hasMore is false.
Inputs: entryId (required), limit, cursor.
get_entry_updatesGet entry updatesARead-onlyIdempotentInspect
Get one entry's typed update history. Mirrors the REST entry-updates endpoint. A page can be empty or shorter than limit while more results remain; continue with pagination.nextCursor until pagination.hasMore is false.
Inputs: entryId (required), field, limit, cursor, +1 more.
get_sourceGet monitored pageARead-onlyIdempotentInspect
Get one monitored page in full, including every setting add_source and update_source can write — userPrompt, notificationsEnabled and updateSensitivity — which list_sources does not return. Read a setting through this tool before changing it, and call it again to confirm the change landed. monitoringFrequency is the cadence stored on the page and effectiveMonitoringFrequency the one in effect: checks run on a scheduler tick, so the two differ whenever the stored cadence is the faster of the pair, and on a verified account with an empty balance the effective value is the weekly rotation (10080) while monitoringFrequency keeps the stored setting. nextCheckTime is absent when there is no next check to report: no check is coming at all (a paused or auto-paused page), or the page has never been checked and its first one has not landed — which is what you will read right after add_source, on an active page. updateSensitivity runs from 0 to 1 and decides how much of a page's churn is worth reporting; the response's sensitivityLevel names the band the number falls in: 0-0.33 is low (material facts only, and a price must move 5% to alert), 0.34-0.66 is medium (the default when it was never set; adds supporting detail, 2% on price), 0.67-1 is high (adds small clarifications too, 0.5% on price). lastListRead says how much of the page the last check read: coverage "partial" means it did not cover the list — it read itemsRead of at least pageItems items — so the page holds items this source is not watching. "whole" means the extraction covered the list. It does NOT mean the page is fully stored: itemsRead is what survived the gates below the extractor, so it can sit under pageItems when they refused rows. itemsDropped is how many they refused: itemsRead plus itemsDropped is what the extraction answered, so anything still short of pageItems is what it never answered. That arithmetic needs itemsDropped to be present — a reading that carries itemsRepeated without it predates the field and is a page whose last check is older than it, not a check that refused nothing. itemsDroppedLost is how many of those refusals cost this page a row; itemsDropped minus it is what the pipeline correctly declined — chrome, rows off the monitoring instructions, a repeat, a shape the contract does not name — so a steady itemsDropped with itemsDroppedLost at 0 is a healthy page and not a short one. It is present whenever itemsDropped is, including as 0, so an itemsDropped with no number beside it is a reading older than the field rather than a check that lost nothing. reason "items_dropped" names the check where they refused every one of them, and itemsRepeated the rows folded into a sibling, which is one part of itemsDropped and not extra to it, and one itemsDroppedLost does not count. reason says what fell short: three budgets (item_cap, prompt_cut, selection), short_answer, which is no budget at all — the check simply came back with fewer items than the page holds — and items_dropped, where the check read the page and a gate below the extractor refused every item it found, leaving itemsRead at 0. itemsRead is what the check took INTO the source, so it can never exceed the rows the page produced. itemsRepeated, when present, is how many rows the check read and did not store as items of their own because another item of the same check had the same identity; a repeat is already outside itemsRead, so the two never overlap, and it is one named part of itemsDropped rather than an addition to it. listReadCut answers what lastListRead cannot: lastListRead describes only the last check and reads unmeasured on a check that did not read the list, so a page whose lowered item cap cut a read and that has not changed since shows nothing there. When listReadCut is present the stored items are SOME of what the page shows, and it stays until a whole-page reread that the item cap no longer cuts clears it (it kept rows, ran at a cap that held, and read whole or short_answer, never item_cap, prompt_cut or selection). Its absence is not proof the page is fully stored: a page whose own count includes the header row of a table reads short_answer even when every row was read, so compare list_source_entries with the page; it is absent when no cut is known, which includes a page that is short only because of its own configured cap. time is when the cut was first noted (epoch milliseconds), which is the check that cut the read, except on a page noted by hand before the field existed, where it is the day it was noted; retries counts the whole-page reads spent on it: below 3 a check that finds the page unchanged reads it whole again (a page that changed keeps its diff, and the retry waits), and at 3 the retries stop and the field stays as the report that the items are short.
Inputs: sourceId (required).
get_source_changesGet recent page changesARead-onlyIdempotentInspect
Get item-level updates across monitored pages, including importance, field values, price deltas, and entry context. Pass sourceId to read one page's changes instead of the whole account's. Either way the read is complete: paged to the end with no sourceId it returns every update the account owns, and narrowing by sourceId returns exactly the subset the unfiltered feed would have reached for that page — so the two agree and neither is a partial history. A page can be empty or shorter than limit while more results remain; continue with pagination.nextCursor until pagination.hasMore is false.
Inputs: limit, since, cursor, sourceId, +1 more.
list_source_entriesList a page's itemsARead-onlyIdempotentInspect
List the items one monitored page currently holds — their IDs, titles, URLs, prices and tracked attributes — most recently updated first. The prose an item carries (its summary and author, for a quote, article or posting) is not here: read an item in full with get_entry. This is how to reach an entryId for get_entry_updates and get_entry_changes: an item that has not changed lately appears in no change feed. An empty first page means the page holds no items at all, which is a different answer from having no recent changes. A page can be empty or shorter than limit while more results remain; continue with pagination.nextCursor until pagination.hasMore is false.
Inputs: sourceId (required), limit, cursor.
list_sourcesList monitored pagesARead-onlyIdempotentInspect
List the user's monitored pages with their current monitoring and fetch health. Read lastListRead before trusting a page's items to be all of them: coverage "partial" means that check did not cover the list, and itemsRead/pageItems say how far it got. "whole" means the extraction covered the list. It does NOT mean the page is fully stored: itemsRead is what survived the gates below the extractor, so it can sit under pageItems when they refused rows. itemsDropped is how many they refused: itemsRead plus itemsDropped is what the extraction answered, so anything still short of pageItems is what it never answered. That arithmetic needs itemsDropped to be present — a reading that carries itemsRepeated without it predates the field and is a page whose last check is older than it, not a check that refused nothing. itemsDroppedLost splits that total into the part that cost the page a row and the part the pipeline correctly declined (itemsDropped minus it): do not read a high itemsDropped as a problem on its own, because page chrome and rows off the monitoring instructions land there every check and are the pipeline working. itemsDroppedLost is the number to act on, and it is present whenever itemsDropped is, including as 0 — a measured "all of those refusals were correct", so an itemsDropped with no number beside it is a reading older than the field. reason "items_dropped" names the check where they refused every one of them, and itemsRepeated the rows folded into a sibling, which is one part of itemsDropped and not extra to it, and one the lost count does not include since the row is held under the item that claimed its identity. listReadCut is the part of that story lastListRead cannot keep: lastListRead describes only the last check, so it reads unmeasured on a check that did not read the list, while listReadCut stays on a page whose item cap had to be lowered and cut a read, until a whole-page reread that the item cap no longer cuts clears it (absent when no cut is known, which includes a page that is short only because of its own configured cap; get_source describes time and retries; retries counts the whole-page reads spent, and a check that finds the page unchanged reads it whole again until it reaches 3). nextCheckTime is when the next check is expected, and it is ABSENT when there is none to report rather than a time that will never arrive: a page you paused, or one auto-paused after repeated failures, has no check coming at all, and a page whose first check has not landed yet has none to name. Read status to tell those apart — the second is active. effectiveMonitoringFrequency is the cadence in effect, which is slower than the stored monitoringFrequency when that is faster than the scheduler's tick. It does not carry a page's monitoring instructions or sensitivity — call get_source for one page's full settings, and list_source_entries for the items it holds. A page can be empty or shorter than limit while more results remain; continue with pagination.nextCursor until pagination.hasMore is false.
Inputs: limit, cursor.
pause_sourcePause monitored pageAIdempotentInspect
Pause monitoring for a page. Reversible: call resume_source to start it again.
Inputs: sourceId (required).
relearn_sourceRe-analyze monitored pageAInspect
Re-analyze how a monitored page is structured, so later checks extract it with freshly learned rules. Use when a page was redesigned, or when its items are being extracted wrongly. Safe: it changes extraction rules only and never touches items already collected. Costs one credit when the re-analysis succeeds, nothing when it fails; refused with noCredits at zero balance.
Inputs: sourceId (required).
resume_sourceResume monitored pageAIdempotentInspect
Start monitoring a paused page again — the counterpart to pause_source. Checks restart on the page's normal frequency and spend credits again, one per page read. Calling it on a page that is already monitored is a no-op, reported as alreadyActive. The first check after a long pause compares the page against how it looked when it was paused, so it can report everything that changed in between. The status field reports the page's stored state, not a prediction: on an account with no credits a resume still answers active, and the page then either keeps a weekly check (verified e-mail, page among its first three months on the free rotation) or is parked again as creditPaused when its next check comes due — call get_credit_balance to tell whether checks will actually run.
Inputs: sourceId (required).
update_sourceUpdate monitored pageAIdempotentInspect
Update ordinary settings for a monitored page. Use pause_source to stop monitoring.
Inputs: sourceId (required), name, userPrompt, updateSensitivity, +2 more.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Related MCP Connectors
Watch a place on a web page (a price, stock, a version) or a whole page; get told when it changes.
Watch web pages and get a short summary of what changed. Create monitors and read changes.
Monitor web pages for changes: create monitors, list detected changes and read diffs.
Extract prices, lists, tables, events from static pages and watch them for changes.
Related MCP Servers
- FlicenseNot gradedqualityBmaintenanceEnables tracking webpage content and price changes over time, including diffs, historical records, and alerts. Supports protected pages with automated browser rendering and anti-bot handling.-
- AlicenseNot gradedqualityDmaintenanceEnables tracking web page changes, detecting content modifications with structured diffs and snapshot history through MCP tools.MIT
- AlicenseAqualityBmaintenanceEnables watching competitor Shopify, WooCommerce, and SaaS pricing and changelog pages, returning only what changed since the last run as price, stock, product, plan, and changelog diffs. Runs on Apify with per-event pricing and saved state keys for separate baselines.1108 npmMIT
- FlicenseNot gradedqualityCmaintenanceEnables monitoring of website changes through contracts, with automatic extraction, semantic diffing, and healing, triggering alerts and actions via email, Slack, or Telegram.-
Glama MCP Gateway
Add one secure layer between your agents and this server.