Supermemory

@supermemory search my meeting notes from last week for any action items assigned to the design team

Found 4 documents matching 'meeting notes' from the past 7 days. Extracted action items: • Design team to finalize mobile mockups by Friday (from 2024-01-15 standup) • Review accessibility audit findings and propose fixes (from 2024-01-16 design sync) • Update component library documentation (from 2024-01-17 all-hands) All items were tagged under the 'design-team' container. Two documents mention dependencies on engineering timelines.

This example shows Supermemory's core retrieval strength: semantic search across previously ingested documents. The MCP requires documents to have been added beforehand via Add Document or Add Documents. Search quality depends on how well your team has tagged containers and whether processing completed successfully.

@supermemory ingest https://github.com/our-org/api-docs/blob/main/CHANGELOG.md and tag it under 'release-notes' so we can reference it later

Document ingested successfully. • Document ID: doc_8x2k9p • Source: https://github.com/our-org/api-docs/blob/main/CHANGELOG.md • Container: release-notes • Status: processing Processing typically completes in 10–30 seconds. Once finished, you'll be able to search this changelog or retrieve its extracted chunks. Use Get Document to check processing status.

Add Document accepts public URLs or raw text. The MCP returns a document ID immediately, but extraction and chunking happen asynchronously. If the URL requires authentication or the content is large, processing may fail silently—always verify with Get Document. This example demonstrates the write path that feeds Supermemory's knowledge base.

@supermemory pull the learned profile for our 'customer-feedback' container and draft a one-paragraph summary of recurring themes

Retrieved profile for container 'customer-feedback' (47 documents, last updated 2024-01-18). Recurring themes: Users consistently request faster export functionality and better mobile offline support. Pricing concerns appear in 18% of feedback, primarily from small teams on the free tier. Positive sentiment clusters around the collaboration features and real-time sync. Three high-priority bugs were mentioned multiple times: CSV export truncation, notification delays, and search result ranking.

Get Profile returns Supermemory's learned summary of a container—useful for synthesis tasks. The profile's accuracy depends on document volume and tagging consistency. This example pairs the MCP's retrieval with the AI's reasoning to surface patterns. Note that profiles are computed periodically, not in real time, so very recent documents may not be reflected.

When Supermemory beats a static wiki for support teams

A 6-person support team fields 40 tickets a day across Slack and email. They've built a 200-doc Notion wiki, but agents waste 8 minutes per ticket hunting the right article. Supermemory wins here because the Get Profile tool learns what each container tag (product area, customer tier) actually needs, then surfaces relevant chunks without the agent writing a search query. Ingest your Notion export with Add Documents, tag by product line, and let the profile API return the three chunks that match the ticket context. The trade-off: if your docs change hourly (release notes, incident logs), the profile bucketing lags by a few hours and you'll still need real-time search. If your knowledge base is stable and your team asks the same 20 questions in different words, Supermemory cuts lookup time in half.

Bulk-loading a new hire's reading list in one call

A 12-person engineering team onboards two developers a quarter. Each new hire gets a 60-doc reading list: architecture decisions, runbooks, style guides, and past postmortems. The hiring manager used to Slack-DM links one at a time; now they export the list as a CSV and call Add Documents once with 60 URLs. Supermemory processes them async, returns a batch result with any failed URLs (paywalled Medium posts, broken links), and the new hire's container tag holds the entire corpus by day two. The API key auth means the manager runs this from a Retool dashboard without OAuth friction. The boundary: if your onboarding docs live in Google Drive with per-file permissions, Supermemory can't crawl them—you'll need to export to public URLs or plain text first. For public or pre-exported content, this beats manually copy-pasting into a chat context window.

Mining 40 retro transcripts for recurring blockers

A 9-person product team runs biweekly retros and saves Zoom transcripts to Dropbox. After 20 sprints, the PM wants to know which blockers keep coming up. She uploads all 40 transcripts with Add Documents, tags them by quarter, then calls Get Profile with the 'blocker' container to see the bucketed themes Supermemory extracted. The profile API groups recurring phrases (deploy pipeline, design handoff, flaky tests) without her writing regex or reading 40 docs. This works because retros use consistent language and the corpus is large enough (40 docs) for the profile to find patterns. The limit: if your retros are under 10 docs or highly unstructured (off-topic chatter, no agenda), the profile won't surface useful buckets and you're better off with keyword search. For teams with 30-plus structured retros, this surfaces the top 5 blockers in under a minute.