Build your cloud context library

Choose Cloud context for this guide#

Context Library opens in Private Vault mode. This guide covers the separate Cloud context tab, which lets Todoki and OpenRouter process source text with your consent. For browser-encrypted storage with a passphrase you control and per-request optional cloud AI, follow the Private Vault guide instead.

1. Set the scene#

1. Open Context library in your workspace. Give the context a project-specific name, such as Atlas product launch.
2. Describe your role, project, audience and goal. For example: I lead Atlas, a scheduling app for agencies. Help me answer customer questions about the launch.
3. Choose Guide me and add specific facts: current prices, supported features, owners, important dates and terminology. Mark proposed dates and planned features clearly.
4. Add boundaries: Keep replies concise. Do not promise unshipped features. Ask when information is missing. Include when your facts were last updated.

2. Review files or a folder#

1. Choose Files or folder and select a local folder or individual files.
2. Todoki reads text-based PDFs, Word DOCX, UTF-8 TXT, Markdown, CSV, JSON, VTT and SRT. PDF/DOCX text is extracted in your browser, with a 5 MB file limit and up to 100 PDF pages. Hidden files, common build folders and credential-named files are excluded. Scans need OCR first; password-protected PDFs, old .doc files, ZIPs and images are unsupported. Export Notion pages as Markdown and select the extracted folder; direct Notion page sync is not included.
3. Review the selected filenames and remove anything you do not want processed. Up to 40 sources and 120,000 text characters fit in one import; the guided context also counts as a source.
4. Use clear filenames and include dates in the text. Remove credentials, wallet recovery phrases and sensitive data you do not want processed. Files remain in the browser until you confirm the import.

3. Review the processing quote#

Processing costs $0.25 per started 20,000 characters, with a $0.50 minimum per import. One 120,000-character import is $1.50. The server verifies the size and price. This is separate from meeting capture and subscription billing.

The initial context checkout uses Base Sepolia tokens with no monetary value. Review the network, amount and receiver yourself before approving the wallet payment. There is no automatic renewal. Processing starts only after verified settlement.

Keep the saved request if payment is interrupted. Do not pay again when it is under review. If source processing fails or pauses, select Resume processing; the existing paid import is reused without another processing fee.

4. Use your context#

1. Wait for the context to show Ready. In Ask your second brain, choose up to three ready contexts.
2. Ask a specific question using the terminology in your sources. For example: What can I tell a customer about our Outlook integration?
3. Open the source details under the answer and check the exact quote. Retrieval selects matching passages; a missing answer can mean the facts are absent or your wording did not find them.
4. During a captured meeting, open Live Copilot and expand Your context. Select your ready packs before requesting a reply or asking a question. Recaps still use only the meeting transcript.

Control what Todoki uses#

Contexts are private to their creator in the selected workspace. Other members and workspace admins cannot retrieve them through Todoki’s context APIs. Source text and the generated index are encrypted in Todoki’s database; Todoki’s server can decrypt them. OpenRouter receives source passages for processing and relevant excerpts when you ask a question.

These imports are not uploaded to Walrus. Archive a context to exclude it from future answers; restore it from Archived contexts. Archiving retains the encrypted data and does not retract content already shown in an earlier answer. Updated files are a new, separately quoted import; there is no background folder monitoring.