Your lab's knowledge,
talking back.
SRKB ingests the PDFs you actually read, extracts fine-grained claims, benchmarks and datasets — and lets you reason over them with a citation-grounded AI agent.
14-day free workspace · no credit card · cancel anytime
Upload, parse, deduplicate
Drop PDFs straight into S3. Workers chunk, embed, extract metadata and flag duplicates across your team.
Grounded conversations
Chat with an agent that cites the evidence it used — every claim links back to the chunk, page and paper.
pgvector semantic search
Query across paper titles, overviews, claims, benchmarks, datasets and metrics — filtered by project.
Structured claim graph
Conclusions, directions and thoughts — stored as first-class objects with evidence, confidence, and author.
Bring your own model
Plug in OpenAI, Anthropic, Azure or a compatible endpoint per tenant. Per-user quotas and audit built in.
Multi-tenant by default
Row-level security, per-project roles, audit logs, and PSP-agnostic billing via Stripe / PayPal / Alipay.
A thin UI over a surprisingly deep pipeline
Each action you take fans out into a Celery job on the worker EC2 — the API never blocks on LLMs or embeddings.
Ingest
Upload PDF → S3. A `parse_embed_job` extracts text chunks, builds embeddings, infers paper metadata and writes it into Postgres.
Organize
Attach papers to projects. Layer claims, benchmarks, datasets and metrics with human-in-the-loop review.
Reason
Open an agent session. It streams tokens, tool calls and retrieved citations back to you in real time via SSE.
Stop re-reading papers. Start querying them.
Spin up a workspace in under a minute. Invite your co-authors. Point an agent at your literature.
Create my workspace