Skip to content
Docs

Vercel Plugin for AI Coding Agents

The Vercel plugin gives supported AI coding tools Vercel-specific context, skills, agents, and slash commands. The default installation keeps automation lightweight and activates session-start context only in empty directories and detected Vercel, Next.js, or eve projects.

npx plugins add vercel/vercel-plugin

The plugin installs Vercel context, skills, and a lightweight default hook profile.

OpenCode 2 uses its own plugin command and doesn't need Node.js or Bun:

opencode plugin add github:vercel/vercel-plugin

To connect the Vercel MCP server, run /mcps in OpenCode, select vercel, and sign in. Slash commands use the vercel: prefix, such as /vercel:deploy prod. Run opencode plugin update to get new versions.

After installation, session context is injected automatically only for empty directories and detected Vercel, Next.js, or eve projects. You can invoke skills and commands directly when you want targeted guidance:

/vercel-plugin:build-agents
/vercel-plugin:ai-sdk
/vercel-plugin:deploy prod
ComponentDescription
Ecosystem graphA relational knowledge graph covering every Vercel product, library, CLI, API, and service, with decision matrices and cross-product workflows
33 skillsDeep-dive guidance for specific Vercel products, libraries, and workflows
3 specialist agentsPurpose-built agents for deployment, performance optimization, and AI architecture
4 slash commandsQuick actions for deploying, managing environment variables, bootstrapping projects, and more

The plugin installs into these coding tools. Each tool has its own installation flow.

ToolStatus
ChatGPTSupported
Claude CodeSupported
Codex CLISupported
CursorSupported
GitHub CopilotSupported
GitHub Copilot CLISupported
Grok BuildSupported
Kimi Code CLISupported
KiroSupported
OpenCode 2Supported
VS CodeSupported

After installation, the plugin keeps automatic behavior lightweight. Session-start activation runs only in empty directories and detected Vercel, Next.js, or eve projects, and Vercel skills are not auto-injected on every prompt or every tool call by default.

The skills remain available on demand, and the plugin includes the injection engine for targeted or future opt-in workflows.

  • Session start context injection: Injects a thin Vercel session context plus knowledge-update guidance for empty directories and detected Vercel, Next.js, or eve projects
  • Session start repo profiler: Scans config files and dependencies to set likely-skill hints after the same activation check passes

The plugin includes 33 skills covering the Vercel ecosystem:

SkillCovers
access-protected-vercel-deploymentAuthenticated CLI and browser access to protected preview and production deployments with vc curl, OIDC, and Trusted Sources
ai-gatewayUnified model API, provider routing, failover, cost tracking, and hundreds of models
ai-sdkAI SDK 7, including text and object generation, streaming, tool calling, agents, MCP, and embeddings
authAuthentication integrations for Clerk, Better Auth, Descope, and Auth0 with Marketplace provisioning
bootstrapProject bootstrapping, including linking, environment provisioning, database setup, and first-run tasks
build-agentsDefault guidance for building AI agents with eve, including tools, channels, schedules, and Slack agents
cdn-cachingCache hit rate, stale content, revalidation, per-request cache reasons, and ISR read and write costs
chat-sdkMulti-platform chat bots for Slack, Telegram, Teams, Discord, Google Chat, GitHub, and Linear
create-a-backendChoosing between Functions, Services, containers, Workflow, Queues, and databases for a backend
custom-metricsApplication and business metrics emitted from Functions and queried with vc metrics
deployments-cicdDeployment and CI/CD workflows, including deploy, promote, rollback, --prebuilt, and CI files
domainsDomain search and pricing, registration, project domains, DNS, transfers, and renewals
env-varsEnvironment variable management, including .env files, vercel env, and OIDC tokens
eveThe eve framework for durable AI agents, including channels, sandboxes, schedules, evals, and frontend clients
flags-sdkFeature flags and A/B tests with the Flags SDK, provider adapters, the vercel flags CLI, and Flags Explorer
is-agenticAgent readiness scores for websites and public MCP endpoints with Is Agentic
knowledge-updateKnowledge update guidance for the plugin
marketplaceIntegration discovery and installation with vercel integration add and vercel install, auto-provisioned environment variables, and unified billing
microfrontendsMicrofrontends on Vercel, including path routing, project composition, local development, and troubleshooting
queuesVercel Queues, including durable topics, consumer groups, retries, delays, and idempotency
react-best-practicesReact and Next.js performance guidance across component, data, and rendering patterns
routing-middlewareRequest interception before cache, rewrites, redirects, and personalization for Edge, Node.js, and Bun
runtime-cacheEphemeral per-region key-value cache with tag-based invalidation across Vercel Functions, Routing Middleware, and Builds
vercel-agentVercel Agent in the dashboard and Slack, including code review, investigations, and approved actions
vercel-cliVercel CLI commands for deploy, env, dev, domains, cache management, MCP integration, and Marketplace provisioning with vercel install
vercel-connectVercel Connect guidance for scoped OAuth tokens, third-party services, MCP servers, and eve agent connections
vercel-firewallDDoS protection, WAF rules, rate limiting, bot filtering, and IP controls
vercel-functionsNode.js, Bun, Python, and Rust runtimes on Fluid Compute, streaming, WebSockets, and Cron Jobs
vercel-sandboxEphemeral Firecracker microVMs for running untrusted or AI-generated code safely
vercel-servicesMultiple frontends and backends in one project, with public rewrites and private service bindings
vercel-storageBlob, Global Config, Neon Postgres, Upstash Redis, one-command Marketplace provisioning with vercel install, and migration from sunset packages
verificationEnd-to-end verification across browser, API, data, and response flows
workflowWorkflow SDK, including durable execution, WorkflowAgent, steps, Worlds, and pause or resume flows

The plugin doesn't ship Next.js skills. Next.js includes version-matched docs for agents, and you can add its workflow skills with npx skills add vercel/next.js. See the Next.js AI agents guide.

The plugin includes three specialist agents:

AgentExpertise
deployment-expertCI/CD pipelines, deploy strategies, troubleshooting, environment variables
performance-optimizerCore Web Vitals, rendering strategies, caching, asset optimization
ai-architectAI application design, model selection, streaming architecture, MCP integration

Use slash commands directly in your AI coding tool:

CommandPurpose
/vercel-plugin:bootstrapBootstrap a project with linking, env provisioning, and db setup
/vercel-plugin:deployDeploy to Vercel (preview or production)
/vercel-plugin:envManage environment variables (list, pull, add, remove, diff)
/vercel-plugin:statusView project status, recent deployments, and environment overview

To deploy to production, pass prod as an argument:

/vercel-plugin:deploy prod

Prompt text and bash and tool-call telemetry are not collected.

Telemetry behavior:

  • If VERCEL_PLUGIN_TELEMETRY is unset, the plugin sends a once-per-day dau:active_today event, and a skill:invoked event with the name of each plugin skill your agent loads
  • VERCEL_PLUGIN_TELEMETRY=off disables all telemetry, including both events

To disable telemetry in shells that launch your AI coding tool:

export VERCEL_PLUGIN_TELEMETRY=off
setx VERCEL_PLUGIN_TELEMETRY off

If the plugin is not behaving as expected, enable debug logging with VERCEL_PLUGIN_LOG_LEVEL:

export VERCEL_PLUGIN_LOG_LEVEL=debug

Available log levels:

LevelDescription
offNo logging (default)
summaryHigh-level injection summaries
debugDetailed matching and dedup information
traceFull pipeline traces with timing breakdowns

You can also run the built-in doctor command:

npx vercel-plugin doctor

The doctor command validates manifest parity, checks hook timeout risk, verifies dedup health, and reports skill map issues.

If a skill gives incorrect advice or injection does not fire when expected, file an issue on GitHub. Include:

  • What you were building
  • What the plugin injected, or did not inject. Enable debug logs with VERCEL_PLUGIN_LOG_LEVEL=debug
  • What was wrong about it
Last updated July 29, 2026

Was this helpful?

supported.