How to Fix “No MCP Servers Configured” in Continue
Fix Continue’s empty MCP server list by checking Agent mode, the active config, server file structure, and launch settings. Find the right logs if a configured server will not start.
If Continue shows “No MCP Servers configured,” first check whether the server entry is being detected at all. Switch to Agent mode, confirm you edited the configuration Continue is actually using, and check that the server is under the mcpServers key. If an entry appears but fails to connect, inspect its launch settings and Continue’s logs. An empty configuration section does not by itself mean a configured server failed at runtime.
1. Check Agent mode and the active configuration
Continue’s MCP integrations are available in Agent mode. Switch to Agent mode before diagnosing the configuration. Then confirm which configuration is active in your Continue client; editing a different config file will not add a server to the active one.
For a workspace-level setup, Continue documents the .continue/mcpServers directory. Check that the file is in the current workspace’s directory and that Continue is opening that workspace. Continue can read YAML and JSON MCP configuration files from this directory.
2. Check the server list structure
A standalone YAML block file needs top-level name, version, and schema metadata, followed by an mcpServers list. Each server entry needs a name. For a locally launched process, it also needs a valid command.
name: Local MCP servers
version: 0.0.1
schema: v1
mcpServers:
- name: Browser search
command: npx
args:
- "@playwright/mcp@latest"
This is a structural example based on Continue’s documented quick start. It does not guarantee the example package is installed or runnable in your environment. Follow the server’s own setup instructions for required packages and arguments.
Check indentation carefully: mcpServers should be a top-level property, and the dash marks each item in the list. If you use JSON, preserve the same configuration fields in valid JSON syntax.
3. Check process launch settings
If the server is listed but does not start, compare its entry with the server’s own documentation. Continue’s configuration reference supports these relevant fields:
| Setting | What to check |
|---|---|
name |
Required label for the server entry. |
command |
Required for a command-launched server. Confirm the executable exists in Continue’s process environment. |
args |
Optional command arguments, in the order the server expects. |
env |
Optional environment variables. Confirm required credentials or paths are available to the launched process. |
cwd |
Optional working directory. Set it if the server relies on files relative to a particular directory. |
| Connection timeout | Optional timeout setting. Check it if the server needs longer to initialize. |
url and transport |
For remote servers, use the URL and HTTP transport configuration required by the server. |
Continue documents stdio, sse, and streamable-http transport labels. A local process configuration and a remote HTTP server have different launch fields; do not copy a different client’s config unchanged without checking Continue’s syntax and the server’s transport requirements.
4. Inspect logs when a server entry exists
Continue recommends checking logs early when troubleshooting a configured integration. In VS Code, open Developer Tools and inspect the Console. In JetBrains, inspect ~/.continue/logs/core.log. Look for errors that identify a missing executable, malformed arguments, unavailable environment variables, or a transport/connection problem.
On macOS, Continue documents spawn ENAMETOOLONG as a process launch issue for which using the full executable path can help. Resolve the executable path with which, then use that path for command, for example:
which npx
Replace npx in the config with the returned path. This workaround applies to that spawn error; it is not a general fix for a server list that is empty.
5. Separate discovery problems from connection problems
- No entry is shown: verify Agent mode, the active workspace/configuration, the
.continue/mcpServerslocation, required top-level metadata, and themcpServerslist shape. - Entry is shown but fails to start: check
command,args,env,cwd, executable availability, and the server’s transport instructions. - Entry is shown but cannot connect remotely: check the configured URL, HTTP transport type, server availability, and Continue logs.
- The message appears on a Hub configuration page: it may simply indicate that the published config has an empty MCP Servers section. That wording alone does not establish that a runtime connection failed.
Or skip the browser setup
If you need a webpage screenshot as part of an agent workflow, ScreenshotNeo is a screenshot API and MCP server for AI agents, including Claude, Cursor, and any MCP client. It provides take_screenshot, get_page_info, and capture_pdf. Its API can return a PNG, JPEG, WebP, or PDF from one GET request. See the ScreenshotNeo API documentation.
curl -G "https://api.screenshotneo.com/v1/shot" -d access_key=YOUR_API_KEY --data-urlencode url=https://stripe.com -o shot.webp
import requests
r = requests.get(
"https://api.screenshotneo.com/v1/shot",
params={"access_key": "YOUR_API_KEY", "url": "https://stripe.com"},
timeout=90,
)
open("shot.webp", "wb").write(r.content)
const q = new URLSearchParams({ access_key: 'YOUR_API_KEY', url: 'https://stripe.com' });
const res = await fetch(`https://api.screenshotneo.com/v1/shot?${q}`);
if (!res.ok) throw new Error(`Screenshot request failed: ${res.status}`);
const bytes = new Uint8Array(await res.arrayBuffer());
await import('node:fs/promises').then(fs => fs.writeFile('shot.webp', bytes));
Cookie banners, newsletter popups, and chat widgets are removed before the shot. Bot checks, blank pages, and failed loads are never billed. It offers an MCP server for AI agents, and 1,000 screenshots a month are free with no card; paid plans start at $5 for 3,000. Sign up for free ScreenshotNeo access.
Common errors and fixes
| Symptom | Likely cause | Fix |
|---|---|---|
| No servers listed | Wrong mode, inactive config, wrong workspace folder, or malformed server-list structure. | Use Agent mode; check the active configuration and workspace’s .continue/mcpServers; validate metadata and mcpServers. |
| Server listed but process will not launch | Missing executable, bad arguments, unavailable environment, or incorrect working directory. | Run the server’s documented command in the relevant environment and correct its command, args, env, or cwd. |
| Remote server will not connect | Incorrect URL or transport configuration, or a server-side connection issue. | Use Continue’s documented HTTP transport form and verify the endpoint and server instructions; inspect logs. |
spawn ENAMETOOLONG on macOS |
Process spawning with the executable name/path fails in the current setup. | Try the resolved executable path from which as Continue documents, and check the log for further launch details. |
| “No MCP Servers configured” on a Hub page | The published configuration page may have no MCP entries. | Distinguish an empty displayed section from an attempted runtime connection; add an entry to the active config if you intended to configure one. |
Reliability and setup notes
- Keep the server’s own installation and environment requirements alongside the Continue config; a syntactically valid entry can still refer to an unavailable program.
- Use logs to distinguish config discovery from process startup and transport connection failures.
- For remote servers, validate the transport form supported by both Continue and that server. Continue documents SSE and streamable HTTP configuration.
- Continue’s cited MCP documentation does not specify a universal server timeout value or performance benchmark. Use the optional connection timeout only where startup time warrants it, and diagnose slow starts from logs.
FAQ
Can I configure an MCP server with a JSON file?
Yes. Continue documents JSON MCP configuration files in the workspace’s .continue/mcpServers directory. Ensure the file uses the expected Continue structure and valid JSON.
Does the message prove the MCP server is down?
No. It can indicate an empty configuration section. A listed server with launch or connection errors is a separate case; inspect the relevant logs.
Can I use MCP in Continue’s chat mode?
Continue’s MCP guide says MCP is available in Agent mode. Switch to Agent mode when using MCP integrations.
Should I copy another MCP client’s configuration directly?
Check it first. Continue accepts its own server configuration shape, and local stdio and remote HTTP servers require different settings.


