Hi - I’m playing around with some mcp servers and I’m trying to connect to one called octagon for deep research. It seems the mcp connects ok and the query is going through the API key (because it takes money from my account there) however after 30 seconds my pickaxe returns a timeout error, saying it waited 30 seconds and received no response from the client. Obviously it’s going to take more than 30 seconds to get back the results from a deep research - any ideas how I can force the mcp to wait say 10 minutes for a response? I tried to tell it to wait in the prompt, but still received the same timeout error. Thanks!
I get the same timeout issue after 60sec on the reasoning models; it would be good if the timeout period was configurable / extended, given the reasoning / deep research models can take quite long to provide a response.
Hi @lachb & @transitionlevel,
We’ve increased the timeout from 30 seconds to 10 minutes. Let’s see how it goes!
1 Like
Thanks - but now it just gives a “network error” after a similar amount of time working. I tried a few a couple of times, same behaviour - not my internet connection.
Hi @lachb,
Are you using this one https://github.com/OctagonAI/octagon-deep-research-mcp?
Do you have any example prompt?
Seems to work fine on my end:
Ok now I have got it to work once or twice. The network timeout error still comes mostly and it seems to be around 5 minutes after it starts researching.
Yes, its that octagon one. I’m just trying with some of the example prompts they give in the documentation
Yes confirming after a couple more tests ‘network error’ is coming exactly 5 minutes after the pickaxe begins researching.
I am getting similar ‘network error’ for MCPs.
Mind sending the exact conversation? I’ll check why network error occured. I also made some changes, maybe that helped.
Hi still hitting network error after five minutes. I literally just test it with the suggestions in the octagon documentation like “Study the cultural impact of streaming services on traditional media consumption” it works for five minutes then ‘network error’.
I am using an MCP that is hooked to Airtable and should retrieve this data. I’m able to get it to do basic functions (List tables in base) but anything more (“How many records have Gateway Status of ‘Active’?”) and it basically bogs down and gives network error after 5 minutes.
On the regular bot side using o1, I still get this after 60sec for queries it works that long on:
HTTPSConnectionPool(host=‘api.openai.com’, port=443): Read timed out. (read timeout=60)
I am not having any issues with the Octagon MCP
Hi - Any updates here. I just checked again and still hit network error bang on 5 minutes after it starts working. Thanks