Nexlayer is the production cloud your coding agent deploys to. It runs your whole application (frontend, API, databases, workers, long-running agents) as one always-on app from one nexlayer.yaml, and gives your agent the tools to deploy, inspect, debug and operate it.
If your agent can build it, Nexlayer can run it.
This repository is where coding agents and people report problems with the Nexlayer MCP server, the Nexlayer CLI and the platform. Nexlayer is a managed, proprietary platform; its source code is not here.
- Docs: https://nexlayer.com/docs
- Written for agents: https://nexlayer.com/llms.txt · full reference: https://nexlayer.com/llms-full.txt
- Dashboard: https://app.nexlayer.com
Pick one. Both sign the person in once, in the browser.
MCP server (Claude Code, Cursor, VS Code, Windsurf, Zed, Codex, any MCP client):
npx @nexlayer/mcp-installOr connect directly: https://mcp.nexlayer.ai/api/mcp (Streamable HTTP, OAuth on first use). Discovery: https://nexlayer.com/.well-known/mcp
CLI (any agent with a terminal, macOS or Linux):
curl -fsSL https://nexlayer.com/cli/install.sh | bash
nexlayer loginSetup for every agent: https://nexlayer.com/docs/mcp/overview
The skills hold the rules for nexlayer.yaml and the build → validate → deploy → verify order.
nexlayer_get_skills list them
nexlayer_get_skill_content skillName=ship-it-nexlayer before any deploy
nexlayer_get_skill_content skillName=debug-nexlayer when an app is broken
Details: https://nexlayer.com/docs/agents/skills
From the MCP: build, validate, deploy, then check status. From the CLI, one command runs the same steps and verifies the app:
nexlayer ship --jsonEvery --json answer is one line: {kind, status, exit_code, next, …}. Follow the next field.
| Exit code | Meaning | What to do |
|---|---|---|
| 0 | Working | Tell the person it's live. |
| 1 | Deployed but not working | Read cause and next, fix it, ship again. |
| 2 | Couldn't confirm | Name the routes that matter with --check, then verify again. |
| 3 | Needs the person | Stop and ask: a key, a card, Docker, sign-in. Never work around it. |
| 4 | Failed | Fix what detail says, ship again. |
| 5 | Nexlayer had a problem | Wait, then retry. Change nothing. |
| 64 | The command line was wrong | See nexlayer help --json. |
CLI reference: https://nexlayer.com/docs/cli
- The app is created by its first deploy and gets a generated URL (or its custom domain from
url:). Redeploys update it in place. - Images are
linux/amd64and carry a tag (neverlatest). - Never put a secret value in
nexlayer.yaml, the repo or the chat. Write a${NAME}reference; the value comes fromnexlayer vault set NAMEor the app's Secrets tab in the dashboard. - Pods reach each other at
<pod-name>.pod:<port>. A pod without apathis private to the app. - Live schema:
nexlayer_get_schema(MCP) ornexlayer call get_schema(CLI). Reference: https://nexlayer.com/docs/deployments/configuration
| Plan | Price | |
|---|---|---|
| Free | $0, card on file, never charged | Up to 5 apps |
| Pro | $100/mo flat | Unlimited apps, unlimited custom domains |
| Enterprise | Contact sales | Multiple workspaces and teams, SSO, RBAC, audit logging, SLAs |
No per-request, per-build or per-token meter and no egress fees. Current prices: https://nexlayer.com/pricing (JSON: https://nexlayer.com/api/pricing)
AI models run on Nexlayer Inference, a separate product: https://nexlayer.com/docs/inference
Agents: call the MCP tool nexlayer_submit_issue. It files the issue here with the ai-agent label and your agent, model and date. Include the tool or command, the app name and the exact error. Never include secret values.
People: open an issue and pick MCP, CLI, Dashboard, Deploys or Inference.
Questions: Community Slack. Security reports: see SECURITY.md; never in a public issue.
- Quickstart: connect your agent, ask it to ship, get a live URL
- Templates: full-stack apps your agent builds from one spec
- Open source: deploy any GitHub repo or open-weight model
- Use cases
- Changelog