Dieter_be —
@user as a prevention against sneaky tiny-font/invisible instructions in emails etc, would it make sense to render the html of an email and do OCR?
#general
Hello everyone, I am participating in the 2fast2mcp hackathon.And it was my hackathon, and i will try do my best .
🙌1
👋 Hi everyone!
@user
Hi Matvey,
What i built actually? Any specific problem statement there?
Hi everyone !!
👋 Hi everyone!, Im participating in hackathon really excited
Hello everyone! I’m really excited to participating in the 2fast2mcp hackathon. Let's all do our best!
Hey everyone, I'm a final year CS undergrad and a backend developer, really excited to be a part of this hackathon and this community as well. Hoping to contribute and add some value. :high_brightness:
👋 Hello, Everyone! I’m forming a team for the hackathon and looking for motivated members.
If you’re interested in collaborating, please message me. Let’s create something impactful.
🙌1
Hello everyone I am new to this hackathon i didnt know what to do and how to start can anyone suggest an idea what to do
hey folks, this is yamini. looking forward to the hackathon
hi, pls share necessary installation and how to enter hackathon for first timers.
Hi everyone! Hope this post may help spark your imagination 🔥
🙌5
Hey, my name is nocde, and I want to know that what exactly we have to build is it mcp servers , agent or what ? need some proper explanation
hey everyone i am new here and i am looking to join a team
Hi! I'm Sanjay, participating in the 2 Fast 2 MCP hackathon (Speed Racer track).
I have a questions:
1. Does Archestra provide any API keys (Anthropic/OpenAI) for hackathon participants, or do we need to
bring our own?
🙌1
2 replies
I guess we've to bring our own
@user Hey Matvey How are you 🙂 do you have any new ai projects that you are currently working one please let me know
🙌1
1 reply
i am running the Archestra locally can we create the API key from this is it free will it work?
the docs didn't mentioned that API keys will be provided rather it says API keys are supported that means you have to provide your API maybe not sure....
✅1
4 replies
Finally its time for archestra hackathon, will shortlist some ideas now
👀1
what do we need to build , like by using archestra we have to build mcp , right??
Hi folks! First time working with MCP here. Excited to be part of the hackathon and learn along the way
🙌1🚀1
shortlisted some crazy ideas! time to implement them, hope so Me and my team is able to complete in time, they're kinda complex but thats where the fun is
🙌1🚀2👏1
Hi guys participating as a solo dev. can someone help me understand how to submit the project.
What am i missing ?
docker pull archestra/platform:latest;
docker run -p 9000:9000 -p 3000:3000 \
-e ARCHESTRA_QUICKSTART=true \
-v /var/run/docker.sock:/var/run/docker.sock \
-v archestra-postgres-data:/var/lib/postgresql/data \
-v archestra-app-data:/app/data \
archestra/platform;
Is it possible to make Archestra modifications for my needs, if necessary? 😉
what time is the live session happening today?
Hey @user @user when and where is the live stream?
Website says right now
Hello everyone, this is Bhavith. I’m excited to be here and looking forward to the hackathon!
🙌1
in yt, it says at 22:30
@user
Oh, ok lemme check
Yes it's 10:30 PM IST
where is the youtube link?
Thanks! @user
👋 Hi everyone! I'm Devanshu and I was actually looking into main repository. there's so much interesting stuff going on.
and so many things to learn. really love it.
Hey everyone I am ayush khairnar excited for hackathon
Hey guys!
I am Madhu, Happy to connect with all of you
will be learning, exploring and building few interesting things and moving to prod faster
are we getting the link to the livestream now?
here?
anyone know that how can i add members in my team?
@user as a prevention against sneaky tiny-font/invisible instructions in emails etc, would it make sense to render the html of an email and do OCR?
:halo_wave:1
9 replies
One last Question guys, Do we have to buid our own UI for the hackathon or can we leverage the Archestra's UI ?
Answer to this question will help to scope the project for the hackthon.
New release! 🚀🚀🚀
1.0.41 (2026-02-09)
Features
- add OAuth 2.1 authorization server for MCP Gateway (#2639) (7d170c1)
- implement prompt draft persistence (#2131) (7ba9963)
- MCP auth at tool call time (#2662) (c75c4f3)
- select model and key per agent (#2626) (fb00f22)
- yaml advanced configuration (#2584) (c23061a)
Bug Fixes
- browser preview improvements (#2580) (6307b90)
- consistent mcp session id (#2661) (97f18ce)
- enable browser mcp (#2689) (e4c6307)
- fix bulding arm platform image (#2638) (cf2065e)
- fix cves (#2591) (580bca1)
- fix mcp browser reinstall (#2660) (b17d764)
- fix streamable default yaml generation (#2601) (16f055d)
- fix subagents browser tabs isolation (#2653) (70950c6)
- improve browser preview (#2585) (112e2e0)
- improve browser preview4 (#2595) (d2acfbc)
- increase stop count from 20 to 500 (#2624) (6b0320a)
- polish advanced yaml configuration (#2598) (4edd98e)
- prepend /v1 for Ollama OpenAI-compat proxy paths (#2617) (3e28b76)
- prevent browser stream unsubscribe from killing in-flight chat t… (#2685) (a6d2a3e)
🔥1
- prevent tool policy dialog header from overflowing boundary (#2602) (8a6bddc)
- rename Langgraph to LangChain in architecture diagram (#2632) (b36dbc7)
- replace alert with sonner toast (#2572) (17e0e1b)
- Revert "Revert "fix: sync CORS and trusted origins, improve origin error UX"" (#2668) (90a7f70)
- running migrations when ext vault secret is used (#2633) (8bf7276)
- set ARCHESTRAORCHESTRATORK8SNODEHOST and lint issue (#2694) (10c4b38)
- surface real error instead of NoOutputGeneratedError in A2A delegation (#2686) (556d56f)
- sync CORS and trusted origins, improve origin error UX (#2656) (c263876)
- update Slack community invite link (#2667) (2234935)
- use semver for version comparison to prevent false "new version" display (#2592) (e2ac086)
- wait for secretmanager to init before run migration (#2678) (6d75487)
Dependencies
- bump @anthropic-ai/sdk from 0.71.2 to 0.72.0 in /platform (#2622) (e854e20)
- bump the platform-dependencies group across 1 directory with 33 updates (#2645) (434a0ee)
Miscellaneous Chores
What a release!
Great release!
Also @user Amazing live stream, cleared many doubts, thank you! Much appreciated
Read-only live mirror of Archestra.AI Slack
👋Join the discussion withAI enthusiasts!@user as a prevention against sneaky tiny-font/invisible instructions in emails etc, would it make sense to render the html of an email and do OCR?
:halo_wave:1
9 replies
Hi @user, nice to see you!
👋 hey !
My 2 cents is that it's definitely an interesting idea.
I think this hidden instructions anyway will be passed to the LLM input and archestra should detect them when trying to evaluate an unathorized tool call, but extra layer of protection before sounds solid anyway.
white text on a white background is not the biggest issue, there may be just a plain text prompt injection in the middle of 60-pages doc 😞
But idea is nice!
nice to see you Dieter 🙂
yeah multi-modal security is.. tricky, If you're bringing it up as a hackathon idea - I think it's a great/ambitious idea!
It's just where my brain went when I saw the "look how easily matvey can hack clawdbot" post
especially since OCR is a solved problem now with several models doing this well (I think ?)
The Dual LLM pattern, w/ OCR as you mentioned, might be one way to approach this here: