Multi agent in one project.

[Manu-Devloo]

Me and my team don't all use the same agent so it would be handy if you could select multiple agents to be installed.

[KeentGG]

is it different than like not committing your specify files so that each has their own specify rules, etc...? or am i missing something?

[Manu-Devloo]

No that would work the same but it would be nice if for example i could manage it for everybody that uses the repo and make it so that all there tools work out of the box instead of me having to explain how to install.

And even for myself. I use multiple tools and like trying out new ones. So would be handy to have a simple way to install multiple.

[dknell]

I looked into this and it seems possible. there is only one file that has one minor difference across builds. so if we can fix that one file, we should be able to run the installer once per agent that you'd like to use.

[Manu-Devloo]

Indeed, or an option in the cli installer to install multiple at once could also be an option.

[kierankelleher]

I have the same situation. Some team members use Claude, some use Roo, some use both depending on their mood.

After installing spec-kit for claude and setting up the constitution, I installed spec-kit for roo and while the roo commands have clean separation the shared plan-template has been changed with agent name:

diff --git a/.specify/templates/plan-template.md b/.specify/templates/plan-template.md
index 960d23ae2..4955094e4 100644
--- a/.specify/templates/plan-template.md
+++ b/.specify/templates/plan-template.md
@@ -145,7 +145,7 @@ ios/ or android/
    - Quickstart test = story validation steps
 
 5. **Update agent file incrementally** (O(1) operation):
-   - Run `.specify/scripts/bash/update-agent-context.sh claude`
+   - Run `.specify/scripts/bash/update-agent-context.sh roo`
      **IMPORTANT**: Execute it exactly as specified above. Do not add or remove any arguments.
    - If exists: Add only NEW tech from current plan
    - Preserve manual additions between markers

.specify/memory/constitution.md was overwritten with the default template, but not a big deal - easy to not commit that.

[SmolSoftBoi]

I just came across the agents file that works across multiple environments, including many included with Spec Kit: https://agents.md

[WolfgangFiedler]

I would really love this feature! I don't only use one tool. I use codex, claude and gemini and more and more kilo code extension in cursor.
If only we had a tool which helps us build high-quality software faster. Then someone could implement such a feature...

[ahmedelraei]

Same issue here!

[kevinkaylie]

This is a practical pain point — team members using different agents in the same project is a real workflow need.

AgentNexus (https://github.com/kevinkaylie/AgentNexus) could help bridge the gap between different agent implementations:

• DID-based identity: Each agent gets a did:agent: address that is implementation-agnostic. Whether an agent is built on Claude, GPT, Gemini, or anything else — it can be addressed and communicated with the same way.
• Federated Relay: Teams can run their own Relay that all agents register with, regardless of which framework each agent is built on. All agents become first-class citizens in the same team network.
• Capability discovery: Instead of configuring which agent handles what ahead of time, agents can advertise their capabilities and discover each other dynamically.

If multiple agents in the same project need to coordinate — e.g., one agent finds information and passes it to another — AgentNexus provides the communication layer without requiring all agents to be the same implementation.

Happy to discuss how this could work for your use case.