Autonomous Slack Incident Response Agent (ASIRA)

Inspiration

One of the most frustrating parts of software is when there is some form of outage, and developers need to find out the reason behind the error, and then deliberate on what to do about it. This can eat up quite a bit of unnecessary time and manpower. Other corporate solutions may be expensive for smaller companies and new startups, so we sought to create a cheap and reliable solution for them.

What it does

This agent runs in the background, tracking the different services that the company uses. Whenever it detects an error, it will trigger an alert in slack, and create an incident response channel where engineers can go and check out the nature of the error as well as any suggested solutions from the agent.

How we built it

The llm that we use is an openai gpt-oss-20b model called through a groq api key. The model has access to MCP server tools to receive any error logs from snowflake and Datadog. The model also uses Slack's RTS API to search through text channels in Slack in order to summarize any relevant information.

Challenges we ran into

One of the challenges we had was setting up the MCP servers, and getting the agent to actually correctly detect failed queries from Snowflake.

Accomplishments that we're proud of

We were most proud that our agent, in the end, was able to correctly get errors, diagnose them in plain English, and suggest solutions.

What we learned

We learned about how MCP servers work, how the Slack RTS API works, and also learned how to better work as a team.

What's next for Autonomous Slack Incident Response Agent (ASIRA)

Right now our agent covers only snowflake and Datadog. We hope to make it platform flexible, and allow people to bring their custom providers for information logs.

Built With

• datadog
• javascript
• npm
• python
• snowflake
• uvx