> ## 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.

# Spending, tracking and stopping

> What an assistant's calls cost, the limits that stop spending, how tracking recurs, how exports are priced, what a refusal means and how to stop.

No plans and no subscriptions. Every workspace starts free and pays as it goes from credits inside one monthly budget.

* **Free:** Public index, weekly check for one brand, all searches, numbers, and excerpts over the last 90 days, your workspace's own answers and pages in full, and 25 public full texts a day.
* **1¢ (\$0.01) standard / 5¢ (\$0.05) frontier:** Fresh AI checks ([`ask_ai_engines`](/docs/mcp/tools/ask_ai_engines): \$0.01 on ChatGPT, Google AI Mode, Perplexity, and search APIs; \$0.05 on Gemini and Claude), live page fetches ([`fetch_live_pages`](/docs/mcp/tools/fetch_live_pages)), and custom classifications ([`classify_items`](/docs/mcp/tools/classify_items)). New workspaces start with \$5 in credit (500 standard checks). A check is one question asked once on one AI engine or search surface from one place.
* **Archive & extra public full texts:** Public answers older than 90 days cost \$0.001 the first time your workspace counts or lists them (auto-capped inside your monthly budget); public full texts past 25/day cost \$0.005 each.
* First top-up starts at \$5 (later top-ups from \$10), valid 12 months after the last top-up. Only a workspace admin adds credit on [Billing](https://parse.gl/app/settings/billing).

## Prices

Standard checks, live pages, judgments and rows outside your own data past the free ones cost \$0.01; frontier checks on Gemini and Claude cost \$0.05. Reading the last 90 days of what Parse already has is free: searches, numbers, excerpts, your own answers and pages in full, and 25 full texts of public answers or pages a day, then \$0.005 each. A public answer older than 90 days costs \$0.001 the first time your workspace counts or lists it, and is yours after.

| Call | Price |
| - | - |
| [`check_brand`](/docs/mcp/tools/check_brand), [`resolve_ids`](/docs/mcp/tools/resolve_ids), [`query_metrics`](/docs/mcp/tools/query_metrics), [`search_answers`](/docs/mcp/tools/search_answers), [`search_cited_pages`](/docs/mcp/tools/search_cited_pages), [`read_item`](/docs/mcp/tools/read_item), [`get_brief`](/docs/mcp/tools/get_brief), [`log_work`](/docs/mcp/tools/log_work), [`list_work`](/docs/mcp/tools/list_work), [`get_account`](/docs/mcp/tools/get_account), [`update_settings`](/docs/mcp/tools/update_settings) | Free |
| A public answer older than 90 days that a call counts ([`query_metrics`](/docs/mcp/tools/query_metrics), [`search_cited_pages`](/docs/mcp/tools/search_cited_pages), [`classify_items`](/docs/mcp/tools/classify_items), [`search_answers`](/docs/mcp/tools/search_answers) with `view: "diff"`) or lists ([`search_answers`](/docs/mcp/tools/search_answers), [`read_item`](/docs/mcp/tools/read_item), [`export`](/docs/mcp/tools/export)) | \$0.001 the first time the workspace uses it, then its own |
| [`read_item`](/docs/mcp/tools/read_item): an answer's or a page's full text | With credits: the workspace's own answers and pages free, and public ones 25 a day free, then \$0.005 each |
| [`ask_ai_engines`](/docs/mcp/tools/ask_ai_engines): ask now, on any AI engine | \$0.01 a check standard; \$0.05 on Gemini and Claude |
| [`fetch_live_pages`](/docs/mcp/tools/fetch_live_pages) | \$0.01 a page; only pages that load are charged |
| [`classify_items`](/docs/mcp/tools/classify_items) | \$0.01 a unit judged |
| A question tracked daily on both counted engines | 1 question × 2 engines × 30 days = 60 checks, \$0.60 a month |
| A question tracked weekly | 1 question × 2 engines × 4.3 weeks = about 9 checks, about \$0.09 a month |
| A watched brand, topic or page | Free |
| [`ask_ai_engines`](/docs/mcp/tools/ask_ai_engines) with `baseline`: about 20 questions on both engines | About 20 questions × 2 engines = 40 checks, \$0.40; recent answers are reused and not charged |
| The first run of the workspace's own Free brand | Free: Parse pays for it |
| [`export`](/docs/mcp/tools/export) inside the workspace's own scope | Free, up to 1,000,000 rows a month |
| Rows outside it, by [`export`](/docs/mcp/tools/export) or by searches and queries | 10,000 a day and 50,000 a month free, then \$0.01 a row |

Every reply shows its cost in `meta.costCredits`. "Share AI answers with other workspaces", a switch a person sets at the first top-up or in Settings, takes 25% off every check. Example: tracking 30 questions daily is 30 questions × 2 engines × 30 days = 1,800 checks, \$18.00 a month.

## The budget

| Limit | Starts at | Who changes it | When it is reached |
| - | - | - | - |
| The monthly budget (`budget.monthlyUsd`): everything Parse spends for the workspace in a UTC month | \$30 a month; the trial credit until the first top-up | An assistant may lower it; only a person raises it, on [Billing](https://parse.gl/app/settings/billing#budget) | `budget_reached`, and tracking waits (`budget_paused`); it works again when the month starts, or once a person raises it |
| Credit | \$5 to start | A person tops up, or turns on auto top-up | `insufficient_credits` |

An assistant that lowers the budget cannot raise it back, even right away: [`update_settings`](/docs/mcp/tools/update_settings) answers `needs_confirm` with a link for a person. Auto top-up never tops up past the budget. Who in the team may do what is on [Rates and limits](/docs/limits.md#team).

## How paid calls run

Paid calls are [`export`](/docs/mcp/tools/export), [`ask_ai_engines`](/docs/mcp/tools/ask_ai_engines), [`fetch_live_pages`](/docs/mcp/tools/fetch_live_pages), [`classify_items`](/docs/mcp/tools/classify_items), [`track`](/docs/mcp/tools/track).

* A paid call runs on the first call when it fits the month's budget and the credit. The price is held, and what the call really cost is charged.
* A paid call a reply offers in `next` says what it costs and runs as offered.
* `quote: true` returns the price and buys nothing. Its `needsYes` is true when the call would pass the budget, the credit does not cover it or spending is paused, with the reason in `needsYesWhy`. `confirm` is no longer needed.
* [`ask_ai_engines`](/docs/mcp/tools/ask_ai_engines) (questions or a `baseline`), [`classify_items`](/docs/mcp/tools/classify_items), [`fetch_live_pages`](/docs/mcp/tools/fetch_live_pages): an identical call within the same UTC hour counts as one call and buys nothing more. Pass an `idempotencyKey` of your own to buy it again; the same key with other arguments is refused, and buys nothing.
* [`export`](/docs/mcp/tools/export) of rows outside the workspace's own scope works the same way: an identical export within the hour returns the first file and charges nothing more. Its quote prices the rows the call returns, counted, capped by `maxRows` when given.
* A unit that fails is refunded; [`read_item`](/docs/mcp/tools/read_item) on the job's id shows the cost so far.
* Searches and queries over the last 90 days are free. A workspace with credit reads them in full, whatever is left of its budget; one with none reads at the free level, and the reply says why in `meta.levelReason`.

| history mode | What a call older than 90 days does |
| - | - |
| `"auto"` | Default. Reads the deepest even sample worth at most 25% of the active monthly budget (capped at \$5) and reuses samples already bought. |
| `"all"` | Counts every matching past answer within your monthly budget (at \$0.001 each once for answers not yet owned). |
| `"free"` | Buys nothing. Reads only the last 90 days and returns `context.history.rungs` estimating fuller counts. |

## Tracking

```json theme={null}
{
  "tool": "track",
  "args": {
    "targets": [
      "best project management tool for agencies"
    ],
    "cadence": "daily"
  }
}
```

* **Questions** (exact words or `qst_` ids) are asked daily or weekly on the counted engines, or on the surfaces `engines` names. Each engine a question is asked on that day is a check, so weekly buys fewer checks at the same price. The charge recurs until you stop it.
* **Brands and topics** (`brd_` or `top_` ids) are free watches. **Pages** (`pg_` ids or URLs) are watched for their citations free, and read for changes at the workspace's pace, \$0.01 a read that loads; a change is `page_changed` in the brief.
* `places` (up to five country codes or place names) and `language` track a question from another place and in another language. Each place asks the question again, a check on each engine: 2 questions from 3 places, asked daily, are 2 questions × 2 engines × 3 places × 30 days = 360 checks, \$3.60 a month. The quote's `monthlyUsd` shows the figure for each place and for all. A surface that takes no place is asked once a question. A unit that fails is not charged.
* `exclude` and `include` take `top_` or `mkt_` ids out of, or back into, the workspace's scope.
* An assistant starts tracking when the person asked for ongoing tracking, not on its own.

## Baselines

[`ask_ai_engines`](/docs/mcp/tools/ask_ai_engines) with `baseline` takes a `brd_` id or a domain. Parse drafts about 20 buyer questions and asks each on the counted engines. It returns a job, and [`read_item`](/docs/mcp/tools/read_item) on the job's id gives the result. For the workspace's own Free brand, the first run is free; `again: true` runs it again, charged.

## Exports

[`export`](/docs/mcp/tools/export) writes the rows of [`search_answers`](/docs/mcp/tools/search_answers), [`search_cited_pages`](/docs/mcp/tools/search_cited_pages), [`query_metrics`](/docs/mcp/tools/query_metrics) (with any of their rows, `by: ["question_group"]` included) to csv or ndjson, up to 20,000 rows a file.

* It needs a first card top-up, and a scope: a brand, market, topic, question, page, site, a subject, or `filter.scope` of `own` or `tracked`.
* The link needs the workspace key and lasts 24 hours, so it does not open in a browser. An assistant with a fetch tool fetches the file and hands it over. One without reads it with [`read_item`](/docs/mcp/tools/read_item) and its `exp_` id, page by page, in its reply, at no further charge: the rows were counted when the file was written. Only the workspace's own export can be read that way.
* A file of at most 3,000 tokens also comes back whole in `data.content`.
* Files carry ids, judgments, counts and snippets; no full texts and no contact routes.
* Parse keeps a record of each export for 24 months, including after the workspace is deleted.

## Refusals about money and limits

| Code | Meaning | What to do |
| - | - | - |
| `needs_confirm` | An assistant asked to raise the budget or resume spending | A person raises the budget or resumes spending on Billing (`approveUrl` in the reply) |
| `budget_reached` | This month's budget would be passed | Wait for `resetsAt`, or a person raises the budget |
| `insufficient_credits` | Not enough credit; nothing was charged | A person tops up on Billing |
| `not_in_scope` | An export without a scope, or before a first top-up | Name a scope; a person tops up once |
| `limit_wait_exceeded` | Too many calls | Wait `retryAfterSeconds` |

Every error code, with the others, is in the [error table](/docs/mcp-reference.md#errors). A refused call charges nothing: its reply says `charged: 0`.

## Stopping

* **A tracked item:** [`track`](/docs/mcp/tools/track) with `"stop": ["trk_…"]`; its history stays. [`get_account`](/docs/mcp/tools/get_account) lists tracked items, including those unread for 30 days with their monthly cost.
* **The brief by email or webhook:** [`update_settings`](/docs/mcp/tools/update_settings) with `brief.channels`, or `brief.schedule: "off"`.
* **An assistant:** a workspace admin revokes its key on [Agents](https://parse.gl/app/settings/agents); it stops within a minute. Or remove Parse in the assistant's own settings. Claude Code: `claude mcp remove parse-agent`.
* **Auto top-up:** turn it off in Settings.
* **A subscription bought before credits:** Settings › Plan.
* **The workspace and its data:** write to [support@parse.gl](mailto:support@parse.gl). See [Data and trust](/docs/trust.md#deleting).

## Fair-use and anti-scraping limits

* Reads inside your workspace's own brands, their markets, and tracked items are never subject to outside-scope row caps.
* Rows outside your workspace's own scope (across exports, searches, and queries) are free up to 10,000 rows/day and 50,000 rows/month, across up to 250 markets/day and 1,000 markets/month.
* Past those allowances, additional outside-scope rows cost \$0.01/row within your monthly budget (otherwise quoted without buying). Trial-only workspaces unlock outside-scope purchases and bulk exports after a first card top-up.
* Usage against these caps appears in `context.allowance` and [`get_account`](/docs/mcp/tools/get_account) (`outsideScope`).


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