Airweave is an open-source dev tool that lets agents search any app. It connects to apps, databases, or document stores and turns their contents into searchable knowledge bases for agents.
A few months ago we were building agents that interacted with various apps. We were frustrated when they struggled with vague natural language requests like "resolve that Jira ticket about missing auth configs", "refund payments in Stripe for unsatisfied customers", or "what were Q1 returns from the financial sheet in gdrive?" The agents would then inefficiently chain multiple function calls, fail to retrieve data, or hallucinate answers.
We also noticed that despite the rise of MCP creating more desire for agents to interact with external resources, the majority of agent dev tooling focused on function calling and actions instead of search. We were annoyed by the lack of tooling that enabled agents to semantically search workspace or database contents, so we started building Airweave first as an internal solution. Then we decided to open-source it and pursue it full time after we saw the positive reactions from agent builders.
The best part for us about building Airweave is continually seeing the amazing things our users create with it.
Congrats on the launch. Can you tell me more about how the underlying data sync works? Does it run on a schedule? What if there are conflicting data? How “real-time” is the data getting indexed?
thanks @renebrandel! ofc, data is synced either manually or scheduled using cron jobs. event-based updates are a pro-feature that also depend on whether the source's API support webhooks. And we use content hashing for version tracking of inserts, updates, and deletions. Lmk if you have any other questions!
@renebrandel Maturity of the API is really important here. If the API allows easy changelogs/webhooks, it can be near real-time. Some more legacy APIs we can tune to 5 min.
It’s great that the team is building this! The product looks really cool and I love that it’s an open-source solution that empowers developers to transform any app's data into a searchable knowledge base.
Airweave
A few months ago we were building agents that interacted with various apps. We were frustrated when they struggled with vague natural language requests like "resolve that Jira ticket about missing auth configs", "refund payments in Stripe for unsatisfied customers", or "what were Q1 returns from the financial sheet in gdrive?" The agents would then inefficiently chain multiple function calls, fail to retrieve data, or hallucinate answers.
We also noticed that despite the rise of MCP creating more desire for agents to interact with external resources, the majority of agent dev tooling focused on function calling and actions instead of search. We were annoyed by the lack of tooling that enabled agents to semantically search workspace or database contents, so we started building Airweave first as an internal solution. Then we decided to open-source it and pursue it full time after we saw the positive reactions from agent builders.
The best part for us about building Airweave is continually seeing the amazing things our users create with it.
Looking forward to the comments 👋
- Lennert & Rauf
Den
You guys are building the most important layer for the next generation of mcp-native applications. 🦾🫡
Airweave
appreciate that @justin_lee27!! 🙏
Airweave
@justin_lee27 Love working with you team!!
AWS Amplify
Congrats on the launch. Can you tell me more about how the underlying data sync works? Does it run on a schedule? What if there are conflicting data? How “real-time” is the data getting indexed?
Airweave
thanks @renebrandel! ofc, data is synced either manually or scheduled using cron jobs. event-based updates are a pro-feature that also depend on whether the source's API support webhooks. And we use content hashing for version tracking of inserts, updates, and deletions. Lmk if you have any other questions!
Airweave
@renebrandel Maturity of the API is really important here. If the API allows easy changelogs/webhooks, it can be near real-time. Some more legacy APIs we can tune to 5 min.