Hmbown/CodeWhale · error · anyhow::Error

Failed to connect MCP server '{server_name}': server is disa

Error message

Failed to connect MCP server '{server_name}': server is disabled

What it means

get_or_connect found the server in static config or dynamic_servers, but McpServerConfig::is_enabled() returned false, so it refuses to spawn/connect. The server exists but was explicitly disabled (an enabled: false style flag in the server entry, or however the config expresses opt-out), and the pool treats that as a hard stop rather than silently connecting.

Source

Thrown at crates/tui/src/mcp.rs:2672

            return self
                .connections
                .get_mut(server_name)
                .ok_or_else(|| anyhow::anyhow!("MCP connection disappeared for {server_name}"));
        }

        self.drop_connection(server_name, "reconnect");

        // Check static config first, then dynamic servers
        let server_config = self
            .config
            .servers
            .get(server_name)
            .cloned()
            .or_else(|| self.dynamic_servers.read().get(server_name).cloned())
            .ok_or_else(|| anyhow::anyhow!("Failed to find MCP server: {server_name}"))?;

        if !server_config.is_enabled() {
            anyhow::bail!("Failed to connect MCP server '{server_name}': server is disabled");
        }

        let mut connection = McpConnection::connect_with_policy(
            server_name.to_string(),
            server_config,
            &self.config.timeouts,
            self.network_policy.as_ref(),
        )
        .await?;
        connection.catalog_generation = self.catalog_generation.load(Ordering::SeqCst);

        self.connections.insert(server_name.to_string(), connection);
        self.connections
            .get_mut(server_name)
            .ok_or_else(|| anyhow::anyhow!("Failed to store MCP connection for {server_name}"))
    }

    /// Connect to all enabled servers, returning errors for failed connections

View on GitHub (pinned to 8880682c63)

Solutions

1. Set the server's enabled flag to true in the MCP config (or remove the disabling override) and save, then retry — the lazy reload picks up the mtime change.
2. Filter call sites on is_enabled() so disabled servers are skipped instead of erroring.
3. If it should stay disabled, remove or stop calling the code path that references it.
4. For dynamic servers, register with the enabled state you actually want.

Example fix

# before (.mcp.json)
{"servers": {"github": {"command": "gh-mcp", "enabled": false}}}

# after
{"servers": {"github": {"command": "gh-mcp", "enabled": true}}}

// Rust: check enablement before dispatch
if let Some(cfg) = pool.config().servers.get(name) {
    if !cfg.is_enabled() {
        bail!("server '{name}' is disabled in config; enable it or skip the call");
    }
}
let conn = pool.get_or_connect(name).await?;

// Rust: catch and route around disabled servers
match pool.get_or_connect(name).await {
    Err(e) if e.to_string().contains("server is disabled") => {
        return Ok(fallback_without(name)); // skip feature relying on this server
    }
    o => o,
}

When it happens

Trigger: Any access path (tool call, resource read, prompt get, connect_all) naming a server whose config sets the enabled flag to false; also dynamic registrations created in a disabled state.

Common situations: Users temporarily disabling a broken server in config and forgetting to re-enable; config templates that ship servers disabled by default; programmatic callers enumerating all keys in the file without filtering on enabled; flip-flopping the flag while a session holds old state.

Related errors

• invalid JSON payload at key {MCP_SERVER_DEFINITIONS_KEY}; co
• invalid MCP server definition list in key {MCP_SERVER_DEFINI
• Failed to write MCP config {}: {}
• MCP config path cannot be empty
• MCP config path cannot contain '..' components

AI-assisted analysis of Hmbown/CodeWhale@8880682c63 (2026-08-16). Data as JSON: /api/errors/bd020ca3328f77d1. Report an issue: GitHub.