[GH-ISSUE #3637] [FEAT]: Plans for A2A? #2348

Closed
opened 2026-02-22 18:29:17 -05:00 by yindo · 4 comments
Owner

Originally created by @pcuci on GitHub (Apr 12, 2025).
Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/3637

What would you like to see?

In line with your recent MCP support https://github.com/Mintplex-Labs/anything-llm/pull/3547

https://github.com/google/A2A would allow for inter-agent communication to perhaps enable completing more complex tasks

Consider a scenario where an AnythingLLM user wants to file an expense report for a business trip:​

  • Expense Agent: Handles the creation and submission of expense reports.​
  • Currency Conversion Agent: Provides real-time exchange rates to convert foreign expenses into the user's home currency.​
  • MCP Integration: Allows the Currency Conversion Agent to access up-to-date financial data from external APIs.​
  • A2A Communication: Enables the Expense Agent to request currency conversion from the Currency Conversion Agent and incorporate the results into the expense report.​
  • ADK Utilization (mandatory?): Both agents are developed using ADK, ensuring compatibility and streamlined communication within the A2A network.​

This collaborative approach would allow for dynamic, multi-agent workflows that can adapt to complex user needs.

Originally created by @pcuci on GitHub (Apr 12, 2025). Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/3637 ### What would you like to see? In line with your recent MCP support https://github.com/Mintplex-Labs/anything-llm/pull/3547 https://github.com/google/A2A would allow for inter-agent communication to perhaps enable completing more complex tasks Consider a scenario where an AnythingLLM user wants to file an expense report for a business trip:​ - Expense Agent: Handles the creation and submission of expense reports.​ - Currency Conversion Agent: Provides real-time exchange rates to convert foreign expenses into the user's home currency.​ - MCP Integration: Allows the Currency Conversion Agent to access up-to-date financial data from external APIs.​ - A2A Communication: Enables the Expense Agent to request currency conversion from the Currency Conversion Agent and incorporate the results into the expense report.​ - ADK Utilization (mandatory?): Both agents are developed using ADK, ensuring compatibility and streamlined communication within the A2A network.​ This collaborative approach would allow for dynamic, multi-agent workflows that can adapt to complex user needs.
yindo added the enhancementfeature request labels 2026-02-22 18:29:17 -05:00
yindo closed this issue 2026-02-22 18:29:17 -05:00
Author
Owner

@husan42 commented on GitHub (Apr 14, 2025):

Does using A2a require changing stdio to SSE mode?

@husan42 commented on GitHub (Apr 14, 2025): Does using A2a require changing stdio to SSE mode?
Author
Owner

@pcuci commented on GitHub (Apr 14, 2025):

not sure, but if required, maybe https://github.com/sparfenyuk/mcp-proxy and https://github.com/punkpeye/mcp-proxy could come in handy

@pcuci commented on GitHub (Apr 14, 2025): not sure, but if required, maybe https://github.com/sparfenyuk/mcp-proxy and https://github.com/punkpeye/mcp-proxy could come in handy
Author
Owner

@timothycarambat commented on GitHub (Apr 15, 2025):

I am not opposed to A2A, however, like with MCP I would first like to gauge adoption of the framework/concept because then we have to consider what a flow for integration looks like and how "developer-focused" it needs to be, or if we can make it all UI-based and simple for the average person.

A2A is still super new right now and we can spin off and chase this down instantly until we can determine if it is useful in the core of the app and to the majority of people in a way the regular user can also understand and leverage.

This is also coming from my extremely cursory understanding of A2A, I have not dove deep into it yet to see how we could get it into the app anyway, so I need to also research it more on my end to see the "ah ha" moment for this.

@timothycarambat commented on GitHub (Apr 15, 2025): I am not opposed to A2A, however, like with MCP I would first like to gauge adoption of the framework/concept because then we have to consider what a flow for integration looks like and how "developer-focused" it needs to be, or if we can make it all UI-based and simple for the average person. A2A is still super new right now and we can spin off and chase this down instantly until we can determine if it is useful in the core of the app and to the majority of people in a way the regular user can also understand and leverage. This is also coming from my extremely cursory understanding of A2A, I have not dove deep into it yet to see how we could get it into the app anyway, so I need to also research it more on my end to see the "ah ha" moment for this.
Author
Owner

@pcuci commented on GitHub (Apr 15, 2025):

while not the same architecture as A2A, the way LangGraph goes about it conceptually might be illustrative? Multi-agent swarms with LangGraph

imagine (in the docker enabled multi-user anything llm context) user 1 creates an agent 1 and attaches some MCP tools to it, and that turns out to be an useful agent. User 1 would probably want to share it, like we do today with custom GPTs in Open AI...

then another user 2 (perhaps in the same organization, or could be federated) creates in turn their own agent 2 and instructs it to rely on agent 1 for a specific set of complex activities that agent 1 already knows how to handle

it seems AnythingLLM is already on the MCP bandwagon, but "tools" are more an intra-Agent functionality. This A2A protocol feels more inter-Agent, so it appears to be complementary at first glance(?)

can't wait to build agents on top of agents, on top of agents, turtles all the way down!

https://google.github.io/A2A/#/topics/a2a_and_mcp

@pcuci commented on GitHub (Apr 15, 2025): while not the same architecture as A2A, the way LangGraph goes about it conceptually might be illustrative? [Multi-agent swarms with LangGraph](https://www.youtube.com/watch?v=JeyDrn1dSUQ) imagine (in the docker enabled multi-user anything llm context) `user 1` creates an `agent 1` and attaches some MCP tools to it, and that turns out to be an useful agent. User 1 would probably want to share it, like we do today with custom GPTs in Open AI... then another `user 2` (perhaps in the same organization, or could be federated) creates in turn their own `agent 2` and instructs it to rely on `agent 1` for a specific set of complex activities that `agent 1` already knows how to handle it seems AnythingLLM is already on the MCP bandwagon, but "tools" are more an intra-Agent functionality. This A2A protocol feels more inter-Agent, so it appears to be complementary at first glance(?) can't wait to build agents on top of agents, on top of agents, turtles all the way down! https://google.github.io/A2A/#/topics/a2a_and_mcp
yindo changed title from [FEAT]: Plans for A2A? to [GH-ISSUE #3637] [FEAT]: Plans for A2A? 2026-06-05 14:46:03 -04:00
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: Mintplex-Labs/anything-llm#2348