Skip to content
ShapelessAI← Back
Brand

Read and write Brand Memory

The account's durable knowledge as files: voice, positioning, product facts, audience, what has worked.

View as Markdown

Read the tree with its completion ring, read or write one file, delete one, export the whole thing as a zip, scan a link and file what it says, or import a document. Every job and agent reads Brand Memory before it writes, so an agent that fixes a fact here fixes every post after it. An agent can complete the onboarding this way: scan the website, then fill in what the scan could not know.

Ask in plain words

  • "read our positioning from Brand Memory"
  • "update the product facts: we now support Bluesky"
  • "scan our website and fill in the brand"
  • "import this PDF about our product into Brand Memory"
  • "export everything you know about my brand"

From an agent host (MCP)

Server: https://shapelessai.com/mcp (add it to your host).

  • brain_tree: Brand Memory's file list with the completion ring. Needs the read scope.
  • brain_read: Read one Brand Memory file. Needs the read scope.
  • brain_write: Write one Brand Memory file (.md paths only) - the agent grounds future posts in it. Needs the write scope.
  • brain_onboarding: What is still open in Brand Memory, as the questions a human answers, with hints. Ask the human, write the answers as markdown under the question headings with brain_write at the item's path, then call again until open is empty. Needs the read scope.

From code (REST)

Bearer API key with the scope each call names (auth, OpenAPI).

  • GET /api/workspace (read scope): Brand Memory's file tree with its completion ring.
  • GET /api/workspace/file (read scope): One Brand Memory file (?path=).
  • PUT /api/workspace/file (write scope): Write one file.
  • DELETE /api/workspace/file (write scope): Delete one file.
  • GET /api/workspace/export (read scope): The whole Brand Memory as a zip.
  • GET /api/workspace/onboarding (read scope): What is still open in Brand Memory, as the questions a human answers, with hints. PUT each answer to /api/workspace/file and call again until open is empty.
  • POST /api/brain/scan (write scope): Read a link and file what it says into Brand Memory (NDJSON).
  • POST /api/brain/import (write scope): Import a .md/.txt/.pdf/.docx document.

From a terminal (CLI)

  • shapeless brain ls
  • shapeless brain onboarding
  • shapeless brain get
  • shapeless brain put
  • shapeless brain rm
  • shapeless brain import
  • shapeless brain export

Example

curl -s "https://shapelessai.com/api/workspace/file?path=positioning.md" -H "Authorization: Bearer $SHAPELESS_API_KEY"
// MCP tool call
{ "tool": "brain_write", "arguments": { "path": "facts.md", "content": "# Facts\n- Publishes to 12 networks, Bluesky included." } }
shapeless brain put facts.md < facts.md

Refusals

StatusMeaning
400Bad path, or frontmatter Brand Memory refuses (an authority that is not rule or reference, a stale_after that is not a real date); the error says which. Fix the file, do not resend it as is.
413Over 256 KB; nothing was written.
502Brand Memory is not answering; nothing was written. Resend later.
400No multipart form, no file, an empty file, or a name we cannot store.
413Over 1 MB of text, over 15 MB of PDF or Word, or extracted text over Brand Memory's 256 KB; nothing was written.
415Not .md, .txt, .pdf or .docx; images go to PUT /api/brain/assets/{filename}.
422No readable text in it (a scanned PDF).

Where else

More in Brand

Use the media library · Know the connected accounts

Every capability · Docs · /llms.txt