> ## Documentation Index
> Fetch the complete documentation index at: /docs/llms.txt
> Use this file to discover all available pages before exploring further.

> ## Agent Instructions
> Assistants connect to Parse over MCP at [https://agent.parse.gl/mcp](https://agent.parse.gl/mcp) (Streamable HTTP). With no sign-in it answers from Parse's public view (the numbers parse.gl's public pages show, up to 5 rows a list), starting with `check_brand`. Signed in with Parse, or sent a workspace key (`parse_mcp_…`), it works in the person's workspace, starting with `get_brief`. Scripts call the same tools over REST at `https://agent.parse.gl/v1/tools/<tool>`.
>
> ```bash
> npx parse-gl init
> ```
>
> To install just the skills: `npx skills add Parsegl/skills`.
> Every tool is also a POST to https://agent.parse.gl/v1/tools/<tool>; its OpenAPI description is https://agent.parse.gl/openapi.json.

# Agencies and local businesses

> How an agency sets up workspaces for clients and compares their brands, how exports are scoped, and how a local business works with places, listings and local questions.

A workspace is the unit of credit, keys, limits and privacy. This page is for working across several brands, and for working by place. For one brand, start with [Your first session](/docs/first-session.md).

## One workspace per client, or one shared

Credits, the monthly budget, keys, saved topics and the record of what was bought belong to a workspace. Choose by what has to be separate:

| | A workspace for each client | One workspace for all |
| - | - | - |
| Credits and limits | Each client's spend has its own balance, caps and receipt | One balance and one set of caps for everything |
| Keys | A key reaches one workspace, so a client's key reaches only that client | Every key reaches every brand |
| Bought answers | Private to that workspace | In one place: nothing to copy between clients |
| Comparing brands | Brand by brand; `query_metrics` reads the public answers about any brand in either setup | `get_brief` with `moves: true` ranks every brand at once |

The starting credit of \$5 is given once for each person. A person adds credit to each workspace, on its [Billing](https://parse.gl/app/settings/billing) page.

## The free slot, credits and tracking

* [`track`](/docs/mcp/tools/track) with `own` makes a business the workspace's brand. Only for the person's own business, after asking; never for a client or a rival. The first brand is free, and Parse pays for its first 20 buyer questions.
* A second brand of your own is not tracked free: one brand per workspace is. Daily tracking of a brand from credits is a person's switch on the brand's row on Billing; an assistant never turns it on.
* [`track`](/docs/mcp/tools/track) watches any brand, topic or page for free, whoever owns it, and asks questions you name for \$0.01 a check. A check is one question asked once on one AI engine or search surface from one place. A question asked daily on both counted engines is 1 question × 2 engines × 30 days = 60 checks, \$0.60 a month. This is how an agency follows a client's brand it does not own. The questions it asks are paid within the workspace's monthly budget.
* A lookup of a prospect with [`resolve_ids`](/docs/mcp/tools/resolve_ids) and the other public tools creates and tracks nothing.

## Compare brands

* **One client's brief:** [`get_brief`](/docs/mcp/tools/get_brief) with `brand` set to the client's `brd_` id scopes the brief to it, asks nobody which brand is theirs, and claims nothing. The event log, [`get_brief`](/docs/mcp/tools/get_brief) with `view: "events"`, takes `brand` too, and `since` and `until` days, because the workspace holds other clients' events.
* **Which of my brands moved most:** [`get_brief`](/docs/mcp/tools/get_brief) with `moves: true`, free. Every brand the workspace has, its own and those it watches, in its own topics, this week against the week before on matched questions. Significant moves first, then by size. A row says `no_move`, `not_measured` or `no_topics` when there is nothing to rank.
* **These four against each other:** [`query_metrics`](/docs/mcp/tools/query_metrics) with `by: ["brand"]` and `brands`, two to 25 ids, so the brands are read in the same answers.
* **Who leads a market:** the `concentration` measure gives the share of answers naming the top 1, 3 and 5 brands.
* Read [Reading the numbers](/docs/numbers.md) before quoting a difference to a client.

## Exports and scope

* [`export`](/docs/mcp/tools/export) writes rows to a file for a client report. It needs a first card top-up and a scope: a brand, market, topic, question, page, site or subject, or `filter.scope` of `own` or `tracked`.
* Rows inside the workspace's own scope (its saved, claimed and tracked brands, the markets they are in, the questions and topics it tracks, its own answers) are free, up to 1,000,000 rows a month. Rows outside it cost \$0.01 a row past the free ones, quoted before the file is written.
* An assistant without a fetch tool reads the file with [`read_item`](/docs/mcp/tools/read_item) and its `exp_` id, page by page, free.
* A file carries ids, judgments, counts, snippets and the question's wording, never full texts or contact routes. Do not resell the rows; see [What the data is for](/docs/limits.md#terms).

## Local businesses: places

A business with several locations is one brand with its locations under it. Each location has its own area and numbers, and they roll up to the brand.

* **Place ids.** [`resolve_ids`](/docs/mcp/tools/resolve_ids) with `kinds: ["place"]` turns "Austin, TX" into a `plc_` id, with its level and its parent. A local brand's row lists its locations and area. An assistant never sees a location code.
* **`filter.place`** is the place a question is about: `plc_` ids, and the places under each (a region holds its cities). `"all"` reads every place.
* **`filter.country`** is where a question was asked from: ISO country codes. An answer asked with no place set counts as US. Use `by: ["place"]` or `by: ["country"]` to cut by either.
* **A local brand's default.** [`query_metrics`](/docs/mcp/tools/query_metrics) with a local subject and no place reads the brand's area: each location's area, one place at a time, as the app reads it. `context.scope` names the places with their ids.
* **The listing measure.** `listing_share` is, of the answers that list local businesses (a numbered list or place cards), the share whose list includes the brand, and its mean rank there. A business no extracted mention names links to no brand, so the share can read low, never high; `unlinked` says how many.
* **Not measured here yet.** A `filter.place` that matches no stored answer returns `data.placeCoverage` with `status: "not_measured_here_yet"`, how many answers the same filter holds one level up, and a `next` call that asks the question from that place. That ask is paid and the call says so. A zero is never printed for a place with no data.
* **Asking from a place.** [`ask_ai_engines`](/docs/mcp/tools/ask_ai_engines) takes `places`: up to five country codes or place names, as someone there would ask. A surface that takes no place is asked once. [`track`](/docs/mcp/tools/track) takes `places` (up to five) and `language`: each place is its own tracked question, priced as one, and the quote shows the monthly cost per place and in all. A baseline by place is not built yet.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.