llms.txt, pricing.md and OpenAPI

Some readers of this site are programs: an assistant asked "what does ChannelForge cost?", a code generator building a client, a crawler indexing the docs. For them ChannelForge publishes a few plain files. Each is generated from the same code the product runs on — the prices from the plans checkout charges, the API description from the routes — so they cannot drift from what the product does.

The files#

All of them are public and need no key. They are cached for up to an hour (the OAuth documents for five minutes).

llms.txt#

llms.txt follows the llmstxt.org convention: a title, a one-paragraph summary, then sections of links an assistant can follow.

• A short description of ChannelForge: what it makes, the platforms it posts to, how work is paid for in credits, and that assistants can run a workspace through the MCP server or the REST API with the workspace's own consent and scopes — and that neither passes on platform analytics.
• Docs: the API reference, the OpenAPI document, the MCP server and the webhooks.
• Documentation: every page of these docs, section by section, each with its one-line description.
• Pricing: a link to pricing.md and every plan in one line, with its price and credits.
• Optional: the pricing page, security, the showcase and sign-up.

Point an assistant at it when you want it to answer questions about ChannelForge from the source rather than from memory.

pricing.md#

pricing.md is the pricing page in Markdown, built from the very plan definitions checkout uses:

• How credits are charged — the price before every job, holds and real charges, what is free (an episode the critic holds back, a failure on our side), the Trial, the weekly coach and comment themes, films' fixed prices, and how the "typically" episode counts are worked out.
• Plans — one table: monthly and yearly prices, credits, projects, channels, seats, GPU jobs at once, API and MCP, premium quality, long-form films, standard episodes typically, Insights and the coach, outlier radar accounts, brand voice and agency mode; then a section per plan with its features and platforms.
• What each plan includes — every feature and the plans that have it.
• What things cost — the typical credits of each action on each plan, from ChannelForge's own measured costs (measured 2026-11-01); estimates are marked as such.
• Credit packs — 500 credits for $55, 1,500 credits for $150, 5,000 credits for $450, 15,000 credits for $1,200.
• For agents and developers — the API reference, the OpenAPI document, the MCP server and llms.txt.

Prices are in US dollars, before tax. One credit is $0.10 of list value. The page and checkout never disagree, because both read the same definitions.

OpenAPI 3.1#

curl -s https://forge.storeswiki.com/api/v1/openapi.json -o channelforge-openapi.json

The document describes every route of the REST API v1 — its scope (in each operation's description), parameters, request bodies and answers — the one error shape, the errors each route can answer, and the bearer key scheme. Its webhooks section describes every event ChannelForge can send, with the signed envelope, its headers and the retry schedule. It is served with Access-Control-Allow-Origin: *, so browser-based tools can read it directly.

What you can do with it:

• Explore: import it into Postman or Insomnia (import from a link, with the address above), set the bearer token to your key, and every route is ready to send.
• Generate types: for TypeScript, npx openapi-typescript https://forge.storeswiki.com/api/v1/openapi.json -o channelforge.d.ts gives typed paths and bodies for any fetch wrapper.
• Generate a client: for Python, openapi-python-client generate --url https://forge.storeswiki.com/api/v1/openapi.json; the OpenAPI Generator covers most other languages.
• Test: validate your mocks and contract tests against the same document ChannelForge is checked against.

Routes that belong to a feature a workspace does not have (films, Insights, storyboards, revisions) are in the document too; for such a workspace they answer 404, as if they did not exist (or 403 where its plan lacks the feature).

The sign-in documents#

MCP clients find out how to sign in by themselves:

• /.well-known/oauth-protected-resource/api/mcp (and the same at /.well-known/oauth-protected-resource) names the MCP endpoint as the resource, ChannelForge as its authorisation server, the four scopes (read, write, jobs, publish) and where the documentation is.
• /.well-known/oauth-authorization-server lists the endpoints for authorising, tokens, registration and revocation, the code flow with PKCE (S256), refresh tokens, Client ID Metadata Documents and Dynamic Client Registration.

See MCP server for what each step asks.

Sitemap and robots#

/sitemap.xml lists the public site: the marketing pages, the legal documents and every documentation page. /robots.txt lets crawlers read all of it and keeps them out of workspaces, account pages, the admin console, the API, private review links, referral redirects and the sign-in steps. No page of a workspace is in the sitemap: the ones meant to be shared — a channel's link page, a media kit's link, a review link — are reached by their own address.

Related#

• Developer overview — every way into ChannelForge.
• REST API reference — the same routes the OpenAPI document describes, for people.
• Credits and plans and Pricing — what pricing.md says, for people.