Keep building in Claude Code, Codex or Cursor. Use Let.com’s Model Context Protocol tools to create projects, deploy source, inspect logs and manage releases from an authorized AI session.
Your editor stays yours. Your hosting stays connected.An agent needs structured actions and useful results, not just instructions in a README. Explore the actual tool names behind a project’s deployment lifecycle.
create_project({
"name": "customer-portal"
}){
"project": {
"id": "example_project",
"name": "customer-portal"
}
}Project creation gives the agent a stable project ID for subsequent actions. Deploying services can introduce usage charges.
A chat client and a coding terminal have different access to your files. Use the deployment tool that matches the agent’s environment.
deploy_filesPass a small application as inline files when the client does not have filesystem access. The result identifies the selected deployment target; inspect deployment status before treating the app as ready.
Remote or local tool sessiondeploy_directoryThe local MCP transport can package a project folder. Ignore rules are respected, and .env files are excluded. Set deployment variables separately rather than bundling secrets with source.
Local transport with folder accesscreate_upload → deploy_uploadAn agent with a shell can prepare an archive, request an upload URL, and deploy the uploaded source. This separates transferring the code from asking the platform to build it.
Upload workflow for source bundlesMCP connects your AI client to hosting capabilities. Authorization decides which projects and actions that connection may use. Client setup and consent depend on the selected integration.
Use server-side variables for credentials. Audit entries identify changes and variable keys without copying secret values into the audit record.
Read the MCP architecture overview ↗Free, Pro and Team include GitHub and MCP deployment. Compare monthly capacity and choose your purchase term.
Start building. Put something real on the web.
No base plan charge
Free to start. No upfront plan charge.
Commercial projects welcome.
More room for production apps and a small team.
Save $336 · regular 48-month total $672
$336 paid upfront for 48 months.
Renews at $336 / 48 months.
Shared projects. More resources. Unlimited members.
Save $600 · regular 48-month total $1,200
$600 paid upfront for 48 months.
Renews at $600 / 48 months.
USD, before tax. Resource allowances reset monthly. Monthly equivalents are rounded for display; the upfront total is exact. Term discounts apply at every renewal. The 48-month term includes 24 free months.
Paid-plan usage budgets cover on-demand usage beyond included resources and add-ons; they exclude the base plan charge. Pro defaults to $50; Team to $100. Budget controls are not an instant hard-stop guarantee. Free has a $0 hard stop. These plans include edge and build allowances. Container compute and database services are metered separately.
| Your plan, in detail | Free | Pro | Team |
|---|---|---|---|
| Monthly equivalent | $0 | $7 | $12.50 |
| Upfront plan total | $0 | $336 | $600 |
| Renewal total | $0 / 48 months | $336 / 48 months | $600 / 48 months |
| Requests | 1M | 10M | 25M |
| Bandwidth | 100 GB | 1 TB | 2.5 TB |
| Edge CPU time | 10M ms | 100M ms | 250M ms |
| Build minutes | 100 | 500 | 1,250 |
| Team members | Solo | Up to 3 | Unlimited |
| Default usage budget | $0 · hard stop | $50 | $100 |
| Commercial use | Included | Included | Included |
| No per-seat charges | Included | Included | Included |
| AI website builder | Included | Included | Included |
| 5,000+ themes | Included | Included | Included |
| GitHub integration | Included | Included | Included |
| MCP integration | Included | Included | Included |
| Global CDN | Included | Included | Included |
| DDoS protection | Included | Included | Included |
TOOLS FOR THE WHOLE PROJECT
An authorized MCP connection can do more than send application files. Supported tools describe databases, persistent storage, container deployments and background work. The available operations depend on the connection and its backend capabilities, so inspect the tool set before planning an automated change.
Create managed Postgres or Redis-compatible Valkey in the project. Postgres provides private access, daily backups and pooled or direct connection references. Valkey uses authentication, private access and AOF persistence. These are the managed engines this tool exposes.
Run a compatible container image and attach durable storage when the service needs it. Review the image, resource size, port and mount path. A new volume is mounted after redeployment; adding it does not rewrite the running process immediately.
Create durable jobs with retry, delay and dead-letter behavior. Choose push delivery to a service or a pull consumer where supported. Set concurrency around the downstream service’s capacity and make the application safe to retry.
CONFIGURATION WITHOUT COPY-PASTED SECRETS
Database connections and other private values can be attached through variable references. The agent can connect resources without receiving the raw database credential in the tool response. Secret variables are sealed and write-only; logging them would defeat the purpose of that boundary.
Use set_variables to attach the required resource references to the intended service. Choose the pooled Postgres connection for application queries and direct access for migrations or session-specific features. Redeploy so running containers resolve the updated configuration.
Read the returned resource state and deployment status rather than assuming that a successful tool call means the app is healthy. Check the service’s actual critical flow. Keep raw credentials out of chat history, source archives and diagnostic messages.
add_database({
"project_id": "your_project_id",
"engine": "postgres",
"name": "db",
"size": "S",
"storage_gb": 1
})Illustrative request · creates a billable resource when executed in an authorized project.
AUTOMATION WITH CONTEXT
Give an assistant a specific project, a concrete desired result and the limits it should respect. Resource creation can introduce charges. Recovery, deletion and permissions should remain explicit decisions, even when the tool interface makes the mechanics straightforward.
Ask the agent to inspect existing services and databases before adding another one. Reuse the intended resource and confirm the target environment. An idempotent workflow should avoid creating duplicates when a tool call is retried after an uncertain response.
Managed Postgres and Valkey are explicit choices. Do not interpret an arbitrary database image as an automatically managed MySQL or MongoDB service. A compatible container alternative needs its own storage, administration and recovery plan.
Inspect database backup status, deployed service state and the application’s response. Optional Postgres HA and point-in-time restore are plan-dependent. Include separately metered containers and databases in the resource review instead of describing all infrastructure as part of the base subscription.
Choose a monthly workload and purchase term. See the included capacity, team seats and modeled delivery costs together.
Static delivery + builds. Commercial use. Zero server compute in this example.
Explore all plans ↗
Let.com Team$12.50/mo$600 paid upfront for 48 months. Same-term renewal: $600.
$11,940–$12,480 over 48 months, billed monthly. Constant usage and today’s rates.
USD before tax. Monthly requests are CDN asset requests, not unique people or page views; GB uses decimal units. Vercel is modeled with Flat Rate CDN enabled for an eligible web application and Basic 2-vCPU builds at $0.007 per minute. Build times are equal in the example, not a performance benchmark. Viewer seats are free on Vercel; the count above is deploying members.
The range reflects unclear credit eligibility for the fixed CDN tier: the lower estimate lets the $20 credit offset CDN plus builds; the upper applies it to builds only. Savings use the lower estimate. Actual invoices depend on configuration. This is a selected Flat Rate CDN configuration, not a claim that it is Vercel’s cheapest possible setup.
Excluded on both sides: server/edge execution, database and container compute, image processing, storage, model/API fees, add-ons and taxes. Let.com’s Edge CPU allowance cannot be equated directly to Vercel’s Fluid Compute CPU and memory billing. Workloads beyond Let.com inclusions are deliberately unpriced here. Multi-year totals assume constant usage and current prices; Vercel is billed monthly, not sold as a four-year commitment.
Rates reviewed October 20, 2026: Vercel Pro + seats · Flat Rate CDN · Basic build rates. Let.com rates come from the plans on this page.
Brand portfolio · planned deployments
Architecture examples for brands in our portfolio. These are planned deployments, not completed migrations or measured case studies.
Audit.com
Audit.com offers security assessments, readiness services and remediation engineering. The proposed Let.com workspace would use managed PostgreSQL for engagement tracking, queued report jobs and asset storage for project files.
Document.com
Document.com combines guided templates, document storage and e-signatures. A planned Let.com backend would keep workflow records in managed PostgreSQL and send document generation to a queue, with files in asset storage.
Message.com
Message.com brings chat, email and phone conversations into one team workspace. A proposed Let.com deployment would pair conversation records in managed PostgreSQL with Valkey for transient state and queued channel events.
This page covers using the Model Context Protocol to control app hosting on Let.com. It is different from hosting your own MCP server: here, an AI client calls Let.com hosting tools through an authorized connection.
The workflow is designed for MCP-capable coding clients, including Claude Code, Codex and Cursor. The available tools depend on transport and granted capabilities. A local folder tool requires local filesystem access; remote clients can use inline files or uploads.
No. Keep your existing coding environment. MCP adds a connection to hosting actions; it does not replace source control, your editor or your framework.
No. MCP has file and upload deployment paths. A connected GitHub repository is a separate workflow for automatic deployments from pushes and pull requests.
Treat the agent’s permissions deliberately. Use variable tools for server configuration and avoid placing secrets in source or browser bundles. The audit trail records changed keys rather than secret values.
Build time depends on the application, and initial authorization can require consent. The existing GitHub repository connection also requires a person in the dashboard. Once authorized, supported MCP tool calls can run from the AI workflow.
Connect the development workflow to hosting, without adding a per-seat fee to your app plan.