Astry delivers an AI‑powered knowledge brain that runs inside a customer’s own cloud, securely indexing data from tools like Slack, Drive, Notion, and Gmail.
Funding
Funding not disclosed
Founders
Product
Problem
Modern enterprises store critical decisions, context, and knowledge across disparate tools such as Slack, Drive, Notion, and Gmail, making it difficult for teams to retrieve relevant information at the moment of decision. Existing AI assistants either ignore this internal knowledge or cannot enforce the organization’s access controls, leading to hallucinations or security risks.
Solution
Astry provides an AI‑driven “brain” that runs inside a customer’s own cloud environment and indexes data from existing collaboration platforms. When a user asks a question, the system builds a per‑request sandbox containing only the files the asker is authorized to see, then runs a large language model on that restricted dataset. Answers are returned in seconds with inline citations linking back to the original source, and all interactions are logged in an append‑only audit trail. The platform continuously updates a structured, entity‑rich wiki and visual graph that surface decisions, contradictions, and relationships, enabling teams to retrieve and verify knowledge without leaving their workflow.
Target Audience
Primary customers are operational teams, product managers, and knowledge workers in mid‑size to large enterprises that need secure, real‑time access to internal knowledge across multiple collaboration tools.
Features
- Native connectors for Slack, Google Drive, Microsoft SharePoint, Notion, Gmail, and Linear with incremental sync and inherited ACLs
- “Ask” interface that delivers sourced answers in real time, filtered by the asker’s permissions and backed by OS‑level sandbox isolation
- Auto‑generated living wiki that extracts entities, decisions, and contradictions, maintaining provenance hashes for each claim
- Interactive knowledge graph visualizing entities and their connections, mirroring the retrieval paths used by the AI
- Per‑request sandboxing and OS‑level isolation ensure the model only accesses data the user is allowed to see, with no parallel permission store
- Full audit log of all queries and decisions, stored in an append‑only, hardware‑sealed format for compliance