Skip to content

fix: issue #418 - check mcp client connections and re-init #419

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
vugenti wants to merge 10 commits into
base: main
Choose a base branch
from
Open

fix: issue #418 - check mcp client connections and re-init #419

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
10 commits
Select commit Hold shift + click to select a range
0c8f17e
fix: issue #418 - check for disrupted mcp client connections and re-i…
vugenti Feb 16, 2026
604b328
issue #418 - fix log message
vugenti Feb 16, 2026
92e37f7
Merge branch 'main' into vugenti/issue418
vugenti Feb 21, 2026
aa2718f
Merge branch 'main' into vugenti/issue418
vugenti Feb 28, 2026
65a3514
issue #418 - refactor mcp connection handling, health-check http conn…
vugenti Mar 2, 2026
5e82c0a
Merge branch 'main' into vugenti/issue418
vugenti Mar 5, 2026
7836331
Merge branch 'main' into vugenti/issue418
vugenti Mar 9, 2026
9d9b832
Merge branch 'main' into vugenti/issue418
vugenti Mar 11, 2026
81056ce
Merge branch 'main' into vugenti/issue418
vugenti Mar 17, 2026
9b1ab9d
Merge branch 'main' into vugenti/issue418
vugenti Mar 18, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 15 additions & 0 deletions core/agent/agent.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -857,6 +857,21 @@ func (a *Agent) consumeJob(job *types.Job, role string) {

availableActions := a.getAvailableActionsForJob(job)
cogitoTools := availableActions.ToCogitoTools(job.GetContext(), a.sharedState)

mcpNeedsInit := false
for _, server := range a.mcpSessions {
err := server.Ping(a.context, &mcp.PingParams{})

@mudler mudler Feb 16, 2026

Copy link
Copy Markdown
Owner

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

while I like this, can we make it restart only the sessions which are unresponsive? it sounds safer to re-start only the MCP session which doesn't reply to Pings

@vugenti vugenti Feb 16, 2026

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

while I like this, can we make it restart only the sessions which are unresponsive? it sounds safer to re-start only the MCP session which doesn't reply to Pings

Agree that would be better but the code to init the mcp actions is a bit tangled and would need a refactor which I don't have time to dig into at the moment given my level of experience with Golang. While I dug into the mcp library for a bit I couldn't find how to quickly re-establish connection in that context. Although again, given my level of experience in Golang this could easily be an oversight on my part.

I'm using this code now and it really solves my problem. Do you have any thoughts on the best way to restructure the initMcpActions to support the re-init of the individual connection? Before investing more time I want to make sure it's in line with your vision for the project.

@mudler mudler Feb 16, 2026

Copy link
Copy Markdown
Owner

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

gotcha, I see. One way would be to simply collect the ones that don't respond to the ping and call Close on these individually. If you have issues just ping me, we can merge this otherwise and I would take over

@vugenti vugenti Feb 21, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I want to make sure that in case of transient connection failure, MCP server restart, or idle timeout that the agent would re-establish a new connection to the MCP server and have all tools available.

It seems like if we were just to call close on the misbehaving transports, we would be stuck with no easy way to re-intialize the session.

I guess the "correct" solution involves refactoring the initMCPActions into multiple methods and when the ping fails, close/remove the broken session and then re-initialize it individually and add it back into the mcpSessions.

Not sure when I would be able to address this but if you are ok merging this and keeping the issue open or having another issue to refactor and improve the efficiency later, I'd be happy to come back around to this at my next opportunity.

@vugenti vugenti Mar 2, 2026

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@mudler I got around to refactoring the http mcp session handling and this is working for me

if err != nil {
mcpNeedsInit = true
xlog.Error("Error pinging MCP server, will re-initialize MCP actions", "error", err)
break
}
}

if mcpNeedsInit {
a.initMCPActions()
}

allActions := append(availableActions, a.mcpActionDefinitions...)

obs := job.Obs
Expand Down
2 changes: 2 additions & 0 deletions core/agent/mcp.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -220,4 +220,6 @@ func (a *Agent) closeMCPServers() {
for _, s := range a.mcpSessions {
s.Close()
}

a.mcpSessions = []*mcp.ClientSession{}
}
Loading