> ## 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.

# Rates and limits

> Free allowances, tracking rates, workspace roles, and the sample floor behind Parse numbers.

Assistants pay as they go from the workspace's credits, with no subscription. No plan beyond Free is sold: daily tracking of a brand is paid from credits. An assistant on `https://agent.parse.gl/mcp` spends within the limits a person sets.

## Free and daily tracking

| | Free | Daily tracking |
| - | - | - |
| Price | \$0 | about \$49 a month a brand, from credits |
| Updates | weekly, on Mondays | daily |
| Engines | ChatGPT | ChatGPT and Google AI Mode |
| Topics of your own | Up to 3 per brand | Up to 15 per brand |
| Card | Not required | Credits |

[Pricing](https://parse.gl/pricing) is the page people buy credits from.

## What Free includes

* See how AI identifies your business, where it mentions it in buyer answers, and what to do next.
* We start with 20 buyer questions about your market. After that, we check for mentions each week.

## Logged out on agent.parse.gl

* Logged out, an assistant reads Parse's public view (5 tools; [the list](/docs/mcp-reference.md#logged-out)), within budgets that keep a first question free and stop bulk copying. A reply a budget stopped says which, when it starts again, and that signing in lifts it.
* One assistant session: 40 calls, 12 different brands and 8 different markets. A first question takes 3 to 7 calls.
* One network address (an IPv6 /64 counts as one): 300 calls and 1,500 rows a UTC day, and 30 sessions. Calls sent without a session (REST, and assistants that keep none) together name 120 different brands and 80 different markets a day from one address; over REST, 500 rows a day.
* Everyone using one assistant app (ChatGPT, Claude, Claude Code, Cursor, the rest together) without signing in: 3,000 calls and 15,000 rows a UTC day.
* Public tool links opened in a browser share a separate daily budget of 1,000 calls and 5,000 rows. One address can use 60 calls and 300 rows. An address can make 50 invalid calls per UTC hour.
* Every reply is a sample of up to 5 rows with no later pages, over the public window, as parse.gl's pages show it. A trend by day, week or month and every list reach back no further than the last 90 days.
* A workspace an agent starts with no account (`POST https://agent.parse.gl/v1/workspaces`) reads within these same budgets: its key adds none. One address starts 10 a UTC day and one assistant app 300; each keeps up to 100 `log_work` calls until a person claims it, and is deleted after 14 days with no call.

## Assistants on agent.parse.gl

* A workspace with no credits: 2,000 calls a day, samples of up to 5 rows, and shares only over groups of 20 or more answers, over the last 90 days and the past answers it bought. A signed-in workspace reads its own brands in full.
* With credits: every row and paging, free over the last 90 days, and full texts (the workspace's own answers and pages free, and public ones 25 a day free, then \$0.005 each). 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. Asks, baselines, live fetches, rubrics and tracking run when called, within the workspace's monthly budget; past it they buy nothing.
* Credits are added by a person: a workspace's first top-up from `$5`, every later one from `$10`. A new workspace starts with `$5` of credit, which comes with the workspace. Credits stay valid 12 months after the last top-up.
* One monthly budget stops spending: everything Parse spends for the workspace in a UTC month counts against it. It starts at `$30` (the trial credit until the first top-up); a person raises it, an assistant may lower it and never raise it. The credit itself is the other stop. What each does when it is reached is on [Spending, tracking and stopping](/docs/credits.md#limits).
* Exports, and pages past the first of a search that names no brand, market, topic, question, page or site, open with a first top-up; a starting credit alone does not open them. Every export names one of those. Rows inside the workspace's own brands, their markets and what it tracks: free, up to 1,000,000 a month. Rows outside, by export or by searches and queries together: 10,000 a day and 50,000 a month free, over replies naming up to 250 markets a day and 1,000 a month, then `$0.01` a row, quoted before anything is bought.
* Who to write to about a cited page: default `read_item` reads omit contacts and are free. Passing `fields: ["contact"]` unlocks actionable contact routes at \$0.05 each (free on misses, bare author names, and 90-day repeats of the same route). Files never carry contact routes.
* The service's own rate limits are 600 calls a minute for a workspace, in bursts of 120; see [Rate limits](/docs/mcp-reference.md#limits). A reply that a limit shortened says so in `limits`; a refusal carries `retryAfterSeconds`. The full price list is on [Pricing](https://parse.gl/pricing.md) and [Spending, tracking and stopping](/docs/credits.md#prices).

## Team and roles

| | Workspace admin | Member | Assistant |
| - | - | - | - |
| Make and revoke keys on Agents | Yes, any key | Yes, their own | No |
| Add credit, save a card, turn on auto top-up (Billing) | Yes | No | No |
| Raise the budget or answer sharing (Settings) | Yes | No; sees them and who can change them | No: it asks for approval |
| Set the brief's webhook address and alerts (Settings) | Yes | No; sees them and who can change them | Yes |
| Lower the budget | Yes | No | Yes, and it cannot raise it back |
| Track the person's own brand (`track` with `own`) | Yes | With the role they have | With a key an admin made, or a sign-in that allowed changes |

* A sign-in reaches the workspace the person picked, with the person's role. A key reaches its workspace and acts as the admin who made it: a claim through a key is checked against that admin's role each time, and a key whose maker is no longer an admin claims nothing.
* Settings shows each change an agent made, with the key that made it. Members see these settings and the names of the admins who can change them.
* To invite someone, open [Members](https://parse.gl/app/settings/members) in Settings: it lists who can open the workspace and who are its admins.

## What the data is for

Data you read through Parse is for your workspace's own work. Don't resell it, don't redistribute its rows in bulk, and don't use it to build a competing dataset or index. Parse may limit a workspace that does to samples. Parse keeps a record of exports for 24 months, including after a workspace is deleted.

## Data limits

* Public figures count up to the last 180 days; each page prints the window it covers.
* Under 20 answers, public pages and the app print a share in lighter ink with no bar; emails withhold the number until there are enough answers.
* Public index questions are asked in English, with the location set to the United States. Parse asks a tracked brand's own topics from its buyers' country, and a local business's own topics name its place. A local business's market questions are asked in each place of its area, as someone there would ask them, on ChatGPT and Google AI Mode on every plan: its home place on Free, up to 15 places on a paid brand.


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