- Brand
- Brizz
- Category
- Data & Analytics
- Primary Subcategory
- Product Analytics & Experimentation
Integration details
Description
Brizz is the product analytics layer for AI agents. It captures every conversation your deployed agents have with their users, evaluates response quality, detects failure patterns, and groups them into issues. Connect your Brizz workspace to ask about your agents in plain language: - Find and read user sessions and their transcripts, errors and tool calls. - Investigate a detected issue: what went wrong, how often, which users it hit, and the evidence behind it. - Query usage, quality, latency and cost metrics, broken down by service, organization, model or cohort. - Manage your workspace: update an issue's status or assignee, link the pull request that fixes it, and edit classifications, cohorts and dashboards. Requires a Brizz account. You choose which workspace to connect when you sign in, and every request is limited to that workspace's data.
- Integration type
- Plugin
- Verification status
- Not applicable
- Platform
- ChatGPT
- Primary Subcategory
- Product Analytics & Experimentation
- Secondary Subcategories
- None listed
- Brand
- Brizz
- Access
- Account required
- First tracked
- 2026-11-20
- Tool count
- 70
- Market
- US
The Primary Subcategory used for this profile’s headline score.
Other Subcategories where the Integration is listed.
Not discoverable in ChatGPT yet
ChatGPT now suggests Plugins on its own when they match a user's request, but not every Plugin is being surfaced yet.We haven’t seen Brizz suggested yet. Sign up and we’ll track it, and your Discoverability Score will appear here once ChatGPT starts surfacing it.
What discovery looks like

Get alerts for Brizz
Get updates when Brizz’s Discoverability Score or category rank changes.
Competing in ChatGPT Product Analytics & Experimentation
View Category70 tools agents can invoke
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Add one processing rule to a service; this call changes no other rule. Preview it first with `brizz_preview_processing_rule` and show the user what it picks. A compaction, background job, handoff, schema load or app message when rule reaches spans arriving within about 10 minutes; a dispatch or app message text rule applies whenever a conversation is built. Returns the rule's id. Each write saves the service's whole rule set, so a change made elsewhere (the dashboard, another agent, or another of your own calls) between this call's read and its save is overwritten. Make these calls one at a time, never in parallel. This tool mutates configuration and requires the `mcp:write` scope.
brizz_add_processing_rule
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Attach an external URL to a Brizz issue — normally the pull request you just opened to fix it — so the work is reachable from the issue panel. Pass a `title`, or the link is labelled with generic text and a reader cannot tell two pull requests apart. Calling this again with a URL already on the issue returns the existing link instead of attaching a second copy. Brizz derives the link's kind (GitHub, Jira, Linear, …) from the URL itself, and attaching the first link moves an untracked issue to `tracked`. Call `brizz_list_issue_links` to see what is already attached. This tool mutates data and requires the `mcp:write` scope.
brizz_add_issue_link
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Group the sessions matching a filter into a distribution — partition by time bucket, model, hasError, sessionOutcome, userSatisfaction, journeyTitle, intentLabel (where the tenant has intent clusters), intent, intentAction, intentObject, intentPromptType (where the tenant has the intents brizz_list_intents lists; pair with an intentPromptType filter for top requests or complaints), issueType, issueSeverity, toolName, toolFailed, repeatedToolSignature, distinctToolCount, terminalErrorClass, hasToolError, orgId, orgName, userId, initiatedBy, timeSinceAnchor and anchorWeek (with one timeRelation filter), customProperty:<key>, or label:<uuid> — and get a count per group. Answers "what's the breakdown of X", "count sessions by Y", or which tool loop explains a slice of an issue, server-side, without streaming every session. Every row counts SESSIONS, so a session that called a tool five times counts once — use `brizz_aggregate_tool_calls` to count the calls themselves. For a per-tool failure rate group by toolName + `toolFailed` (did THAT tool fail in that session); `hasToolError` is session-wide, so toolName + hasToolError answers the different question "sessions that called X and failed at anything" and overcounts X's failures. Pass `metrics` for sum/avg/min/max/percentile over cost, duration, or tokens, with optional `having`/`order_by`. Same filters as `brizz_search_sessions`. `total_groups` counts every group; when it is larger than the rows returned, page on with the `next_page` offset, start and end.
brizz_aggregate_sessions
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Return a grouped distribution over INDIVIDUAL TOOL CALLS — the per-call counterpart to `brizz_aggregate_sessions`, which counts a session that called one tool five times as one. Use it to rank an agent's tools by call volume, compute a real per-call success or error rate, or reconcile a tool-call count with another system. Partition by time bucket, toolName, toolOutcome, errorCategory, agentName, callType, skillKey, or customProperty:<key> (a call without the property takes its session's value, the one the session set outside its LLM calls), and get a call count or duration metrics per group. Counts are tool calls, each once: a call the app ran outside what is traced counts by the model's request for it, and a tool run through a dispatcher such as execute_tool counts under the dispatcher's name unless the tool's own span reported a failure, where `brizz_aggregate_sessions` groups it under both. Filter on toolName, toolOutcome, errorCategory, agentName, callType, sessionId, skillKey, durationSeconds, or a custom property. This ranks and counts — to read the failure messages behind a slice, take the sessions (`brizz_search_sessions` with a toolName filter) into `brizz_get_session_errors`. Tool errors before 2026-07-27 are substantially under-counted, so a window straddling that date shows a jump that is improved detection, not a regression. `total_groups` counts every group; when it is larger than the rows returned, page on with the `next_page` offset, start and end.
brizz_aggregate_tool_calls
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Last-resort analyst that runs a full agent turn to answer an open-ended question about a service — "why are users frustrated", "what's regressing", "why are sessions failing". It orchestrates the other Brizz tools into one synthesized answer, so reach for it only when no typed tool answers the question directly, or when the user explicitly asks for the analyst. Creates a one-shot conversation; use `brizz_continue_agent_conversation` for a dialog. Requires the `mcp:write` scope. Pass any id in full and verbatim: session and trace ids are 32 hex chars, span ids 16, and issue/cluster/journey/label ids 36-char UUIDs.
brizz_ask_agent
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Hold a multi-turn analyst conversation: `method=start` opens one, `method=send` posts a follow-up to an existing `conversation_id`. Use it when a question needs clarification or chained follow-ups; for a single question prefer `brizz_ask_agent`. Requires the `mcp:write` scope. Pass any id in full and verbatim: session and trace ids are 32 hex chars, span ids 16, and issue/cluster/journey/label ids 36-char UUIDs.
brizz_continue_agent_conversation
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Create a new classification: a label set Brizz will apply to a service's sessions from now on, or with scope issue to each new issue. Give it a name, the labels with a description each, and, for a session classification only, optionally a session filter to scope it (e.g. only sessions where a custom property marks a particular surface). The label descriptions are what the model reads to decide, so write them as instructions, not as titles. For a session classification, dry-run the label set against real sessions first with `brizz_preview_classification`; it does not preview against issues. This tool mutates configuration and requires the `mcp:write` scope.
brizz_create_classification
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Save a named user cohort on a service: a set of user conditions (from brizz_list_cohort_filters), all of which must hold. It shows on the Users page's Cohorts tab and narrows sessions through the `cohortId` filter. Membership is evaluated over each query's own time window. An unknown field or operator is refused when the cohort is saved. This tool mutates configuration and requires the `mcp:write` scope.
brizz_create_cohort
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Create a custom dashboard on a service: blank, as an editable copy of a built-in board (`from_builtin`), or as a copy of another custom dashboard (`copy_of`), optionally with cards to add. Custom-chart queries are run once first, so a bad field fails the call and nothing is created. Returns the dashboard's charts and layout; change them with brizz_edit_dashboard. This tool creates configuration and requires the `mcp:write` scope.
brizz_create_dashboard
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Remove one processing rule, named by its id; this call changes no other rule. Pass `confirm: true`; without it the call names the rule and returns it unchanged. The result carries the removed rule, to add back. What it already shaped stays until a session is rebuilt, and a handoff rule's runs already grouped stay grouped. Each write saves the service's whole rule set, so a change made elsewhere (the dashboard, another agent, or another of your own calls) between this call's read and its save is overwritten. Make these calls one at a time, never in parallel. This tool mutates configuration and requires the `mcp:write` scope.
brizz_delete_processing_rule
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Delete a classification and its bindings on every service it covers — not just the one named. Sessions already labelled keep the labels they were given, but nothing new will be classified and the definition is gone. Requires `confirm: true`. To stop a classification without losing it, pause it instead with `brizz_update_classification { enabled: false }`. This tool mutates configuration and requires the `mcp:write` scope.
brizz_delete_classification
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Delete a user cohort. Any chart or filter that narrows to it by `cohortId` stops working, and the definition cannot be recovered. Requires `confirm: true`. This tool mutates configuration and requires the `mcp:write` scope.
brizz_delete_cohort
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Delete a custom dashboard, its charts, their Slack digest subscriptions and its revision history, for everyone in the tenant. Cannot be undone, and requires `confirm: true`. To keep it while changing it, use brizz_edit_dashboard; to start over from it, copy it with brizz_create_dashboard `copy_of` first. This tool mutates configuration and requires the `mcp:write` scope.
brizz_delete_dashboard
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Describe what this tenant can filter, group, and query on: the metric catalog, the issue priority/severity/status enums, the top intent labels where the tenant has them, the available custom-property names with sample values, and which property identifies an end-customer organization. Call it once at the start of an investigation so filter values come from what exists rather than a guess.
brizz_describe_schema
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Run one saved report's SQL fresh and return its contents (CSV or JSON) inline, truncated to a bounded size — enough to inspect the report's shape, not a full export. Call `brizz_list_reports` first to discover a report_id.
brizz_download_report
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Add, reword, or delete the labels inside a classification. Pass `add`, `update`, and `remove` — the tool reads the current label set, applies just those changes, and writes it back, so labels you do not mention are never touched. A label's description is what the model reads when deciding, so rewording one is the main lever for fixing misclassification. For a deliberate full rewrite, pass `replace: true` with the complete `labels` set — that deletes every label not listed. This tool mutates configuration and requires the `mcp:write` scope.
brizz_edit_classification_labels
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Change a custom dashboard: add, update, move, resize or remove cards, rename it, or roll it back to an earlier revision. Operations apply in order and save as one revision. Position is x, y, w, h on the 12-column grid (update_chart with only those fields moves or resizes a card); the grid floats cards up into empty space, as the dashboard does, and the narrower screen layouts are derived from it. Custom-chart queries run once before saving, so a bad field fails the call with nothing saved. Removing a chart needs `confirm: true`, and also deletes any Slack digest subscribed to that chart, which a restore does not bring back. Otherwise every save is a revision that restore_revision can undo (the newest 50 are kept). If someone saves the dashboard in between, the edit is re-applied to their version rather than overwriting it — except restore_revision, which replaces the dashboard as it stands. This tool mutates configuration and requires the `mcp:write` scope.
brizz_edit_dashboard
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Read one processing rule by the id `brizz_list_processing_rules` gave it.
brizz_get_processing_rule
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Read one classification's full definition: every label with the description that steers the model, the guidance prompt, whether a session may carry more than one label, which services and which sessions it runs over, what share of them it samples, and which parts of the conversation it is allowed to read. Also reports how each label is actually being used over the window, plus the count of in-scope sessions (or, for an issue classification, issues) it never classified. Read this before editing — the label descriptions are the part that decides classification quality.
brizz_get_classification
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Read one user cohort: its conditions, who made it, and the session filter that applies it (`session_filter`), ready to pass to brizz_search_sessions or the aggregate tools. Read this before brizz_update_cohort, which addresses conditions by field.
brizz_get_cohort
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Fetch one conversation turn's complete content, addressed by the `id` `brizz_get_conversation` returned for it (a span id also works, returning every item that span produced). Returns the message text, the full tool input and tool result, the model's reasoning text, the structured error detail, and the item's span attributes. This is the drill-down step: read the transcript, pick the turn that looks wrong, then read exactly what it carried. A field larger than `max_bytes` (default 20000) comes back as `{text, length, offset, truncated}` reporting its true size and pages by `offset` — nothing is ever shortened silently.
brizz_get_conversation_item
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Read one custom dashboard: its charts — title, kind, id — and where each sits on the 12-column grid (x, y, w, h), plus its version and link. `detailed` adds each chart's query and the recent revisions, which brizz_edit_dashboard `restore_revision` can roll back to. Read this before editing: brizz_edit_dashboard addresses charts by the titles and ids shown here. A chart group is one card that draws several of the board's charts as tabs, one at a time; each of those charts keeps its own row, kind `in group <title>`, but draws inside the group's card, so moving it changes nothing on screen and removing it takes it out of the group.
brizz_get_dashboard
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Fetch the opening user message — the first thing each user said — for a set of known session ids. Use it after `brizz_search_sessions` to sample what a cohort was actually asking for, or to read the intent behind a group of sessions without pulling transcripts. Takes at most 50 session ids per call; split a larger cohort into several calls. Returns a map of session_id to first user message text.
brizz_get_sessions_first_user_messages
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. The curated Telemetry Health Center document for one service: every check in the catalog — missing session ids, unrecorded tool spans, absent user identity, and the rest — with what the last sweep found, its evidence, and how to fix it. `problem_count` counts failing checks worth acting on; `setup_count` counts failing checks that are capability never wired up rather than something broken; both exclude anything already dismissed or marked fixed. `no_recent_traffic` means the last sweep had nothing to look at, so every count here is drawn from an empty window rather than a clean bill of health. Pass free-text `notes` (max 4000 runes) to steer it toward what you already suspect. Mutate a specific check, or trigger a fresh sweep, with `brizz_update_health_check`.
brizz_health_check
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Read one user intent in depth: its daily prompt trend and the change from the previous window of the same length, how its sessions ended (completed, mixed, abandoned, failed, not engaged) and how satisfied its users were, and the distinct prompts behind it, most repeated first, each with a session id and any open issue. Pass the action, object and prompt_type of a `brizz_list_intents` row. Follow a prompt's session id into `brizz_get_conversation` to read what the agent did with it, and find the intent's issues with `brizz_search_issues` {intent_action, intent_object, intent_prompt_type}.
brizz_get_intent
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Load one investigation runbook's full instructions by exact name from `brizz_list_skills`: the step-by-step tool sequence, what to look for at each step, and how to conclude. Use it whenever the user wants to investigate, diagnose, triage, debug, or root-cause something and a runbook covers it — read the runbook before working the task, then follow it. A runbook that is itself wrong — a step naming something that does not exist, or one you followed to the end without getting closer to an answer — is worth reporting with `brizz_submit_feedback` when the work ends, as a `flow` report.
brizz_get_skill
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Fetch one issue's detail: metadata, its weekly occurrence series, the most recent evidence sessions with the prompt each user opened with, the ids of the newest 50 evidence sessions (newest first), and the activity log. Read `occurrence_series` rather than `trend_status` — the series is the evidence and the status is a derived label that can disagree with it. The per-session label matrix is omitted unless `include_labels: true`, since it describes how the issue was classified rather than what went wrong. To read the actual failure, follow an evidence session id into `brizz_get_conversation`. If you picked this issue up yourself rather than being handed it, report what you do with `brizz_report_issue_progress` so the work shows in the issue panel.
brizz_get_issue
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. The curated "Investigate with AI" handoff document for one issue, assembled in a single read: concept guidance for its issue type, the agent's setup (model, params, tools), whether a system prompt was captured, per-finding evidence, conversation snippets windowed around the flagged turns, and a follow-up funnel. Pass free-text `notes` (max 4000 runes) to steer it with context only you have — a user complaint, a hypothesis, what you already ruled out. It is one heavy read; if it fails, fall back to the individual issue, finding and conversation reads. Size the issue from `total_session_count`, never from the snippets — `rendered_session_count` is only how many transcripts fit the document's budget. If you picked this issue up yourself rather than being handed it, report what you do with `brizz_report_issue_progress` so the work shows in the issue panel. When you open a pull request for this issue, attach it with `brizz_add_issue_link` so the fix is reachable from the issue panel.
brizz_investigate_issue
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. The full captured system prompt(s) behind one issue, one entry per agent — the root agent first, then agents owning a finding, then the rest, alphabetically. Nothing else on this server returns system-prompt text. Pass `agent_name` (or "root") to fetch one agent instead of the whole roster. A prompt commonly runs to tens of thousands of characters, so each entry reports `{text, length, offset, truncated}` and you page the rest with a larger `offset`. Read the prompt from `structuredContent`; the markdown block is a labelled preview of each window.
brizz_get_issue_prompt
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. The headline analytics surface, for one end-customer or the whole tenant. With an `organization` (id, exact name, fuzzy name, or domain) it returns that customer's brief: usage counters, quality metrics, top users, top journeys, top issues, and recent sessions — the dashboard's organization drawer. Without one it returns a tenant-wide overview: top organizations by activity, global totals, and where to drill next. Good for account reviews, QBRs, and customer meeting prep. Multiple name matches come back as a candidate list (`status: ambiguous`) to pick from.
brizz_get_organization_overview
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Return one session's RAW OTel spans as structured JSON — `spanAttributes`, `statusCode`/`statusMessage`, `spanKind`, `parentSpanId`, `traceId`, and timing — rather than the filtered conversation items. Use it to reconstruct non-terminating loops, retry storms, and structural failures the transcript does not model. One 50-span page per call in ingestion order: walk a long session with `offset` (0, 50, 100…) until `returned` drops below 50, and sort a page by `timestamp` for execution order. Bulk prompt/completion attributes are omitted unless `attributes: "full"`. Span events and stack traces are not included — use `brizz_get_session_errors` for those. Pass `span_id` to read one span in full instead of a page: every attribute value windowed rather than capped at 2000 characters — this is what a failing model call's exact serialized request, or a tool call's exact input and output, actually looked like; page a long value with `attribute_offset`.
brizz_get_session_spans
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Fetch the errors recorded across up to 50 known session ids in one query — the cheap way to check whether a cohort shares a failure instead of pulling each transcript. Returns both failure surfaces, tagged with `source`: standalone error items and tool calls that failed. The two sets barely overlap, so checking only one misses most failures. Each error carries type, category, message, stacktrace, span name, and the tool calls that preceded it. Sessions with no errors are reported as checked-and-clean — with `start`/`end`, clean means no errors inside that window — and ids with no spans in this service are reported separately as not found. Pass `start`/`end` when you have a window — it is the only thing that narrows the scan; it does not affect not found.
brizz_get_session_errors
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Fetch named event payloads for a set of known session ids. Events are the custom signals your agent emits — feedback, checkout, onboarding, conversion. Returns a map of session_id to {event_name: latest payload}. Pass specific event names, or ["*"] for all of them. Call `brizz_list_event_names` first to see which events this service emits.
brizz_get_sessions_events
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Fetch one session's summary by id: start and end time, duration, user, organization, tokens, cost, models, journey, issue count, custom properties, and its metric scores. Each score carries its name, value, unit, scope, and source: `brizz` for scores Brizz computes (a category such as session_outcome is printed with its label), `external` for scores the customer reported with record_metric, whose names are whatever that customer sends. A customer score reported per turn is valued by the customer's session-level report when one was sent, otherwise by the mean of its turns. Use it when you already have a session id and want the summary rather than the transcript. For what was actually said and done inside the session, use `brizz_get_conversation`.
brizz_get_session
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. Return the tool definitions a session's calls resolved to: for each tool, the latest definition captured by the time of the call, with its description and parameter/output JSON schema, or the earliest known one when the call predates them all (`version_resolved: earliest`). This may differ from what the model read: several versions of a tool can run side by side, and the newest captured one answers for all of them. A tool that returned to an older definition still resolves to the newer one. A tool run through a dispatcher such as `execute_tool` resolves to the last definition captured for the service by then, which the model may never have read. Use it when a tool was called with wrong or malformed arguments: ambiguous parameter names and missing descriptions explain a large share of "the model used it wrong". A tool redefined mid-session, going by the definitions captured, possibly from another session, is named in `redefined_during_session`: compare it with the failures before concluding. `version_resolved` says how the calls' timing matched the definition, not that the model read it. `parameters_note` marks an entry with no parameter schema: the tool may take none, or capture may have missed them, and `brizz_get_session_spans` shows what a span declared. Tools with no recorded definition are listed under `tools_without_schema`. Pass `tool_name` to narrow to one tool.
brizz_get_session_tool_schemas
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. A one-user usage rollup over a time window: total sessions, first and last active date, average and median session duration, the journeys the user engaged with and how often, and a deduplicated list of their opening prompts. Identify the user through `filters` — by `userId`, or by email as a custom property. Call `brizz_list_session_filters` first for the right field name for your service. Defaults to the last 30 days.
brizz_get_user_usage
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. List the classifications configured for a service — the customer-defined label sets Brizz applies to sessions or issues (intent, outcome, owning team, and any other taxonomy the team maintains). Each row reports its categories, what it is scoped to, whether it is running, and how it is performing over the window: volume, coverage, and impact (the issue rate for a session classification, the sessions its issues reached for an issue classification). `origin: brizz` marks a Brizz-managed classification, which is read-only. Start here before editing a taxonomy; read one definition in full with `brizz_get_classification`.
brizz_list_classifications
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. List every field a cohort condition can use — user activity, identity, organization and more — with the operators and value shape each accepts, plus the service's user custom-property keys. These are user fields, not the session fields brizz_list_session_filters lists. Call this before brizz_create_cohort or brizz_update_cohort.
brizz_list_cohort_filters
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. List a service's user cohorts — name, condition count, source, last update, id. A cohort is a saved set of user conditions; sessions narrow to one with the filter `{field: "cohortId", operator: "equals", value: <id>}` in brizz_search_sessions and the aggregate tools. Read a cohort's conditions with brizz_get_cohort.
brizz_list_cohorts
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. List the custom-property keys attached to a service's sessions — plan_tier, account_id, environment, and whatever else the service sends. They are service-defined, so the list is dynamic and paginated. Filter on one by passing it as a `filters` field to `brizz_search_sessions` with `filter_type` set to "custom_properties" and an operator: include / exclude for an exact list of values such as IDs (case-sensitive; a comma-separated string, or an array when a value itself contains a comma), equals / not_equals for one value or an array of up to 50 matched as any of (case-insensitive), contains / not_contains for a case-insensitive substring (`contains "148"` also matches `14851`), or exists / not_exists for sessions with or without the key, where an empty value still counts as having it (takes no value). include / exclude take at most 1,000 values per condition, and the whole filter has a size limit: for a longer include or exclude list, split it across calls, not across conditions in one call. contains / not_contains take at most 50 values per condition: for a longer contains list of real substrings, run one session search per batch and combine the results; for a longer not_contains list of real substrings, add several not_contains conditions to one call, since conditions are AND-ed. Quote numeric IDs as strings. Cheap to call — no heavy queries.
brizz_list_custom_properties
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. List what a dashboard card can be. Without data_source: the custom-chart styles and data sources, text and section cards, and the few Brizz charts a custom query cannot draw, with their parameters and default sizes. With data_source: the metrics, functions, dimensions and filters a custom chart's query on that source accepts, including this service's custom properties. Build a chart as a `query`; to start from a built-in board's charts, create the dashboard with brizz_create_dashboard `from_builtin`. Use it before brizz_edit_dashboard `add_chart`.
brizz_list_chart_types
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. List a service's custom dashboards — name, chart count, last update, link — and the built-in boards Brizz ships, which brizz_create_dashboard can copy into an editable dashboard. Read a dashboard's charts and layout with brizz_get_dashboard.
brizz_list_dashboards
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. List the event names a service's sessions emit — onboarding, feedback, checkout, and any other custom signal — with occurrence counts. Use a returned name with `brizz_get_sessions_events` to fetch that event's payload per session. Cheap to call.
brizz_list_event_names
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. List the investigation runbooks this server ships — step-by-step Brizz tool sequences for recurring tasks: diagnose a failing session, triage a service, analyze tool failures, chase a cost spike, reproduce a user complaint, close instrumentation gaps. Returns each runbook's name, what it does, and when to use it, so the whole catalog is cheap to hold in context. Start here when a request is open-ended ("what's wrong with my agent", "why did this session fail") rather than improvising a tool sequence; load the steps with `brizz_get_skill`.
brizz_list_skills
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. The per-finding evidence rows behind one issue: recorded reasoning, error_type, analyzer metadata, and session/span pointers. A span-backed row carries a `conversation_item` pointer whose `brizz_get_conversation_item` call retrieves the source turn's full tool result; a session-level finding carries a `conversation` pointer instead. Narrow with `error_type` (one exact value), `search`, or `session_id`; paginate with `limit`/`offset`, max 50 rows. Pass `fields` on the first call rather than after a truncated one — per-finding `metadata` is about two thirds of the bytes and is shed first under the response budget.
brizz_get_issue_findings
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. List the external links attached to a Brizz issue — pull requests, tickets, session recordings — with each one's kind and when it was added. Use it to check whether the fix you are about to link is already recorded, or to find the pull request someone else opened for this issue.
brizz_list_issue_links
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. List the organizations (end-customers) sending traffic for one service, ranked by activity, with usage counters — sessions, users, issues, cost — and lifecycle status. Supports a free-text `search` over organization name and id. Use it to discover which customers exist before drilling into one, or to surface the heaviest accounts. Paginated.
brizz_list_organizations
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. List a service's processing rules, each with its id, and the rule format. Processing rules shape how sessions read: show a summarizing step as a compaction, keep the app's background jobs out of the conversation, mark the user messages the app wrote, unwrap a tool that runs other tools, and group the spans of an agent a tool started, and read the tool definitions a loader tool returns. Build a new rule from the keys `brizz_list_span_attribute_keys` lists, try it with `brizz_preview_processing_rule`, then save it with `brizz_add_processing_rule`; change or remove one by id with `brizz_update_processing_rule` or `brizz_delete_processing_rule`.
brizz_list_processing_rules
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. List the reports (CSV/JSON exports) a user has saved for a service via the agent's report tool, newest first, with each report's id, filename, format, and creation time. Use a returned report_id with `brizz_download_report` to read its contents.
brizz_list_reports
[Session tracking] Results from this server carry a `brizz_mcp_session_id`; pass it back on every later call in this conversation. List the telemetry services (the deployed agents/apps) this tenant monitors, with each service's id, name, whether it is the default, the last UTC day it received events, the UTC start day of its latest session with non-zero LLM cost, and its 7-day session count; `?` means that service's activity could not be read. The default service is listed first. Every other tool resolves the service on its own, so reach for this only when a tool reports a "multiple services" ambiguity or the user wants to compare services. Cheap to call.
brizz_list_services
Brizz ChatGPT Plugin FAQ
How the directory, categories and Discoverability Score work.
Read the methodologyHow do I improve Brizz's ChatGPT Plugin discoverability?
The levers are the listing surface agents actually read: names, descriptions, keywords, tool metadata, and registry health. Which lever matters depends on where discovery breaks, which is what continuous measurement shows.
What are Brizz alternatives on ChatGPT?
As of 2026-11-20, Brizz competes with PostHog, Mixpanel, Amplitude, Amplitude EU, Statsig, Datadog Experiments, ABMeter, Adobe CJA and 39 more in ChatGPT Product Analytics & Experimentation, ranked by public Discoverability Score.
Where is this profile measured?
This profile uses the geography attached to the latest public registry snapshot: US. Locale tags are intentionally omitted.