Setup
Connect your AI assistant
One-time OAuth setup. After that, your agent researches your queue whenever you ask.
AWB is a private research and memory backend for your own AI agent. You capture items here; your assistant researches them through MCP and writes structured findings back to your account.
The setup instructions below are public and readable without JavaScript. Account data stays private and is only available to an assistant after you approve OAuth access.
Plain-text agent setup URL: https://assistwithbuying.com/ai.txt.
A private account is required for OAuth approval
Anyone can read these setup instructions. When an AI client opens the OAuth approval screen, sign in to the free AWB account whose private workspace you want that client to use.
If you are an AI agent reading this page
Do not try to read the user's private research by fetching website pages. The web app intentionally hides account data from ordinary unauthenticated requests.
To work with AWB, the user must connect you as an MCP client using the endpoint https://assistwithbuying.com/mcp. Authentication is OAuth in the user's browser. Do not ask the user for an AWB API key; static API keys are not shipped yet.
AWB currently exposes twenty-two MCP tools, and many AI apps ask the user to approve individual tool calls. Tell the user this is normal: they can approve AWB tools as they come up, or pre-approve/trust the AWB connector if their AI app offers that setting and they are comfortable.
For the verified Claude setup, tell the user to start from Claude on a desktop app, desktop browser, or claude.ai in a mobile browser. Do not tell them to add the custom connector inside the Claude mobile app; it can use an existing connector after setup, but does not currently expose the add flow. Paste only the MCP server URL and leave Advanced settings blank. AWB does not require an OAuth client ID or client secret for this path.
If you only see the old heading "Your AI assistant" and "Loading...", your fetch layer is reading a stale cached copy. Fetch https://assistwithbuying.com/ai.txt instead.
Capture queue items can be pending or needs-attention retries, and can be new items or follow-ups to existing research. Captures may also include a shopping/research sessionId and ownershipStatus. Preserve those when saving research so the user can later filter a session to bought/owned items. If a capture includes follow-up context, call awb_get_research_context before writing updates.
Captures may include selected Research Tasks. Use those researchInterestKeys as item-level research instructions to decide what sections to write, and copy them into the item profile when they describe the user's intent. For follow-ups, treat the note/photos as additional context for the existing item, not as a separate item. If the user selected "Make a sales listing," write a natural marketplace-ready listing as a custom dossier section with sectionKey sales_listing.
Captures and item profiles may include explicit priceValues rows with amount cents, used/new condition, discovery method, market family, similarity, evidence links, notes, and as-of dates. Use those UI-provided values before inferring price from notes; older priceContext rows are fallback compatibility data only. If the user selected "Price guide by place," estimate Yard Sale, Facebook Marketplace, eBay, Antique store, and New exact or New comparable expectations, include a yard sale row even when confidence is low, store the structured rows in itemProfile.priceGuide, and also write a pricing dossier section with sectionKey venue_price_guide. Do not fabricate an exact new price when only comparable replacement pricing exists.
For capture photos, call awb_get_capture_images when there are several file ids, or awb_get_capture_image for one file, and inspect the returned inline image. Do not try web_fetch, bash, curl, or browser fetch on files[].url first; an AI app's ordinary network sandbox may block AWB/Backblaze hosts even though MCP can return the image directly. When one uploaded user photo clearly represents the item, pass its AWB file id as mainImageFileId on awb_save_research_result or awb_upsert_interest, not a web/reference image URL.
When writing research, use a short section summary plus deeper markdown. For follow-up or custom user requests, use a stable sectionKey like user_request:<slug> so more than one custom section can exist and be edited later. Save durable facts as itemClaims, log ruled-out mistakes with awb_log_research_correction, and pass dependsOnClaimKeys on prose sections so stale sections can be flagged when facts change. Pass structuredData as an object or array, not a JSON string, and write plain text/markdown rather than HTML-escaped text. After completing normal queue research, prefer awb_save_research_result so profile, canonical claims, corrections, open questions, sections, source checks, reference images, stats, and capture completion save in one approval.
Recommended setup
Using a different AI app?
The same MCP server URL works in other apps. Paste it wherever the steps below say “the MCP server URL above.”
For agents and developers