Use GliderTrace with Codex

Codex Setup

Configure GliderTrace with Codex CLI, Codex IDE extension, and Codex app.

Codex

Prerequisite: .NET 10 SDK must be installed before running dotnet tool install --global glider-trace.

dotnet tool install --global glider-trace

Codex starts stdio MCP servers through the configured command. Make sure glider-trace is available on PATH.

# bash/zsh
export PATH="$PATH:$HOME/.dotnet/tools"
which glider-trace

Create .codex/config.toml in each worktree root. Codex loads project configuration for trusted projects. Launch the server from that root; --workspace . selects its launch directory. Use ~/.codex/config.toml only for user-wide configuration. Codex MCP docs: developers.openai.com/codex/mcp.

# .codex/config.toml in each worktree
# Launch the MCP server from that worktree root.
[mcp_servers.glider-trace]
command = "glider-trace"
args = ["--workspace", "."]
startup_timeout_sec = 30
tool_timeout_sec = 1200

In the Codex app, open Settings > Integrations & MCP to add or manage the same MCP server configuration. Advanced edits still go through config.toml.

The GliderTrace plugin is published from glidermcp/glidermcp.com and bundles the product skill plus stdio MCP config. After adding the marketplace, install glider-trace from the Codex app Plugin Directory or from the CLI /plugins browser.

codex plugin marketplace add glidermcp/glidermcp.com

If you prefer Streamable HTTP, start GliderTrace yourself and configure Codex with http://localhost:5003/mcp.

glider-trace --transport http --port 5003

# ~/.codex/config.toml
[mcp_servers.glider-trace]
url = "http://localhost:5003/mcp"
tool_timeout_sec = 1200

Start a new Codex session, then run /mcp to confirm the server is available.

codex
# Then run: /mcp

Use a project-local config: Claude Code .mcp.json, Codex .codex/config.toml (trusted projects), Cursor .cursor/mcp.json, or VS Code .vscode/mcp.json. Include portable config files in version control, or copy local config files into each worktree. Git does not copy ignored files. Use absolute paths only when you intentionally want one fixed checkout, such as a global Claude Desktop configuration.

Open the client at the selected worktree root. Relative server arguments use the server launch directory, not the config file directory. Project scope alone does not guarantee that directory. If the client starts elsewhere, configure its MCP working directory or use a Git-root launcher.

Run git rev-parse --show-toplevel in the client terminal, then ask trace_status for the server workspace. Confirm that the loaded solution or workspace belongs to that worktree before you query or change code. For a wrong root, restart GliderTrace with --workspace set to the intended worktree.

An existing HTTP server keeps its own workspace when the client changes directories. Use separate servers and ports for simultaneous worktrees.

git rev-parse --show-toplevel

Use this alternative when the client starts inside the worktree but below its root. It requires Git and a POSIX shell. Replace the existing server entry; do not add a second instance. For Glider, replace src/App.slnx with your solution path. On Windows, use the relative configuration above and set the MCP launch directory to the worktree root. This launcher cannot identify your worktree if the client starts outside its directory tree.

# .codex/config.toml in each worktree
[mcp_servers.glider-trace]
command = "sh"
args = ["-c", "repo_root=$(git rev-parse --show-toplevel) && exec glider-trace --workspace \"$repo_root\""]
startup_timeout_sec = 120
tool_timeout_sec = 1800

After connecting GliderTrace, add this product-specific instruction to your MCP client system prompt or project instructions so the agent chooses the right tool surface first.

When you need runtime evidence for .NET code, prefer glider-trace mcp before ad hoc shell runs. Use GliderTrace for test runs, workspace command runs, failure summaries, exception and stack evidence, stdout/stderr summaries, counters, traces, dumps, GC dumps, and artifact indexing. Use plain CLI commands only for simple file operations or commands where no captured evidence or session history is needed.