
vercel-open-agents
Build and run durable background coding agents with workflow orchestration, isolated sandboxes, and GitHub integration on Vercel.
- Installs
- 165
- Last 24h
- —
- Trending rank
- —
- Security
- Not audited
Install
Works with every agent the skills CLI supports.
npx skills add https://github.com/reason-machines/ai-agent-skills --skill vercel-open-agents- 01
Install
Run the command in your project. The CLI asks which agents to install for.
- 02
Describe the task
Ask your agent for something this skill covers. It loads vercel-open-agents automatically when relevant.
- 03
Review the result
The agent follows the SKILL.md steps, scripts and references bundled with the skill.
SKILL.md
The instructions your agent reads when the skill is triggered.
Vercel Open Agents
Skill by ara.so — AI Agent Skills collection.
Open Agents is an open-source template for building cloud-based coding agents that run as durable workflows on Vercel. It provides web UI, agent runtime, sandbox orchestration, and GitHub integration to execute multi-step coding tasks without local execution.
Architecture
Open Agents uses a three-layer architecture:
Web App (Next.js) → Agent Workflow (Vercel Workflow SDK) → Sandbox VM (isolated execution)
Key principle: The agent runs outside the sandbox and interacts with it through tools. This separation enables:
- Durable execution beyond single request lifecycles
- Independent sandbox hibernation and resume
- Flexible model/provider swapping
- Clean separation of concerns
Installation
Deploy to Vercel
-
Fork and import the repository to Vercel
-
Generate secrets:
openssl rand -base64 32 # for BETTER_AUTH_SECRET -
Set required environment variables:
POSTGRES_URL=your_neon_postgres_url BETTER_AUTH_SECRET=your_generated_secret -
Deploy to get production URL
-
Create Vercel OAuth app at https://vercel.com/account/settings/oauth-apps
- Callback:
https://YOUR_DOMAIN/api/auth/callback/vercel
- Callback:
-
Add OAuth credentials:
NEXT_PUBLIC_VERCEL_APP_CLIENT_ID=your_client_id VERCEL_APP_CLIENT_SECRET=your_client_secret -
Create GitHub App (optional, for repo access):
- Homepage:
https://YOUR_DOMAIN - Callback:
https://YOUR_DOMAIN/api/auth/callback/github - Setup URL:
https://YOUR_DOMAIN/api/github/app/callback - Make app public for org installs
- Homepage:
-
Add GitHub credentials:
NEXT_PUBLIC_GITHUB_CLIENT_ID=your_github_app_client_id GITHUB_CLIENT_SECRET=your_github_app_client_secret GITHUB_APP_ID=your_app_id GITHUB_APP_PRIVATE_KEY=your_pem_private_key NEXT_PUBLIC_GITHUB_APP_SLUG=your_app_slug GITHUB_WEBHOOK_SECRET=your_webhook_secret
Local Development
# Install dependencies
bun install
# Copy environment template
cp apps/web/.env.example apps/web/.env
# Fill in required variables in apps/web/.env
# Start dev server
bun run web
For existing Vercel projects:
vc env pull
Core Concepts
Workflows
Agent runs execute as Vercel Workflow SDK workflows, enabling:
- Durable execution: survives across steps and deployments
- Streaming: real-time updates to the UI
- Cancellation: user-initiated stop
- Resume: reconnect to active runs
Example workflow structure:
// apps/web/app/api/workflows/agent/route.ts
import { createWorkflow } from '@vercel/workflow-sdk';
export const POST = createWorkflow(async (context, data) => {
const { message, sessionId } = data;
// Agent execution happens in workflow steps
const result = await context.run('agent-step', async () => {
return await runAgent(message, sessionId);
});
return result;
});
Sandboxes
Sandboxes provide isolated execution environments with:
- Filesystem access: read, write, search files
- Shell execution: run commands
- Git operations: clone, branch, commit, push
- Port exposure:
3000,5173,4321,8000 - Snapshot-based resume: hibernate and restore state
// packages/sandbox/src/client.ts
import { createSandbox } from '@vercel-labs/open-agents/sandbox';
const sandbox = await createSandbox({
baseSnapshotId: process.env.VERCEL_SANDBOX_BASE_SNAPSHOT_ID,
});
// Execute commands
const result = await sandbox.exec('npm install');
// Read files
const content = await sandbox.readFile('package.json');
// Write files
await sandbox.writeFile('src/config.ts', configContent);
Tools
Agents interact with sandboxes through tools:
// packages/agent/src/tools/file-edit.ts
export const fileEditTool = {
name: 'file_edit',
description: 'Edit a file with search-and-replace',
parameters: z.object({
path: z.string(),
search: z.string(),
replace: z.string(),
}),
execute: async ({ path, search, replace }, context) => {
const content = await context.sandbox.readFile(path);
const updated = content.replace(search, replace);
await context.sandbox.writeFile(path, updated);
return { success: true };
},
};
Available tools:
- file_read: Read file contents
- file_write: Write entire file
- file_edit: Search-and-replace edits
- file_search: Search file contents
- shell: Execute shell commands
- task: Create subtasks for the agent
- web: Fetch web content
- skill: Access domain-specific skills
Creating Custom Tools
- Define the tool schema:
// packages/agent/src/tools/my-tool.ts
import { z } from 'zod';
export const myCustomTool = {
name: 'my_tool',
description: 'Does something specific',
parameters: z.object({
input: z.string().describe('The input parameter'),
}),
execute: async ({ input }, context) => {
// Access sandbox
const result = await context.sandbox.exec(`echo ${input}`);
// Return structured data
return {
output: result.stdout,
exitCode: result.exitCode,
};
},
};
- Register the tool:
// packages/agent/src/tools/index.ts
import { myCustomTool } from './my-tool';
export const tools = [
fileReadTool,
fileWriteTool,
shellTool,
myCustomTool, // Add your tool
// ...
];
Skills System
Skills are domain-specific knowledge modules loaded into agent context:
// packages/agent/src/skills/registry.ts
export interface Skill {
id: string;
name: string;
description: string;
content: string; // Markdown content
triggers: string[];
}
// Skills are cached in Redis/KV or in-memory
const skillsCache = new SkillsCache({
redis: process.env.REDIS_URL,
});
// Retrieve skill by trigger
const skill = await skillsCache.findByTrigger('nextjs app router');
Create a custom skill:
// packages/agent/src/skills/custom-skill.ts
export const customSkill: Skill = {
id: 'custom-framework',
name: 'Custom Framework',
description: 'Expertise in CustomFramework usage',
content: `
# Custom Framework Skill
## Installation
\`\`\`bash
npm install custom-framework
\`\`\`
## Usage
\`\`\`typescript
import { Framework } from 'custom-framework';
const app = new Framework();
app.start();
\`\`\`
`,
triggers: [
'use custom framework',
'setup custom framework',
'configure custom framework',
],
};
GitHub Integration
Auto-commit and PR Creation
Configure agent to automatically commit and create PRs:
// packages/agent/src/workflows/agent-workflow.ts
export async function runAgentWorkflow(
message: string,
options: {
autoCommit?: boolean;
autoPR?: boolean;
repository?: string;
branch?: string;
}
) {
const result = await agent.run(message);
if (options.autoCommit && result.success) {
await sandbox.exec('git add .');
await sandbox.exec(`git commit -m "Agent: ${message}"`);
if (options.autoPR) {
await createPullRequest({
repo: options.repository,
branch: options.branch,
title: `AI Agent: ${message}`,
body: result.summary,
});
}
}
}
GitHub App Installation
// apps/web/lib/github/app.ts
import { App } from '@octokit/app';
const app = new App({
appId: process.env.GITHUB_APP_ID,
privateKey: process.env.GITHUB_APP_PRIVATE_KEY,
});
// Get installation token
const installation = await app.octokit.request(
'GET /users/{username}/installation',
{ username: 'user' }
);
const token = await app.octokit.auth({
type: 'installation',
installationId: installation.data.id,
});
Configuration
Resource Profiles
For Hobby tier deployments:
OPEN_AGENTS_RESOURCE_PROFILE=hobby
This adjusts chat and sandbox resource limits to Hobby-compatible values.
Sandbox Base Snapshot
Use a preconfigured sandbox image:
VERCEL_SANDBOX_BASE_SNAPSHOT_ID=your_snapshot_id
Create base snapshot:
bun run sandbox:snapshot-base
Voice Input
Enable voice transcription with ElevenLabs:
ELEVENLABS_API_KEY=your_api_key
Optional Caching
Enable Redis/KV for skills metadata:
REDIS_URL=your_redis_url
KV_URL=your_kv_url
Common Patterns
Multi-step Agent Task
// packages/agent/src/agent.ts
export async function runMultiStepTask(goal: string) {
const steps = await planSteps(goal);
for (const step of steps) {
const result = await executeStep(step, {
tools: availableTools,
sandbox: currentSandbox,
});
if (!result.success) {
// Handle failure, retry, or replan
return { success: false, error: result.error };
}
}
return { success: true, steps };
}
Streaming Agent Response
// apps/web/app/api/chat/route.ts
import { streamText } from 'ai';
export async function POST(req: Request) {
const { messages, sessionId } = await req.json();
const result = await streamText({
model: openai('gpt-4'),
messages,
tools: {
file_read: fileReadTool,
file_write: fileWriteTool,
shell: shellTool,
},
onFinish: async ({ text, toolCalls }) => {
// Save to session
await saveMessage(sessionId, { text, toolCalls });
},
});
return result.toDataStreamResponse();
}
Session Management
// apps/web/lib/sessions.ts
import { db } from './db';
export async function createSession(userId: string, metadata: object) {
const session = await db.session.create({
data: {
userId,
metadata,
createdAt: new Date(),
},
});
return session;
}
export async function shareSession(sessionId: string) {
const shareToken = await generateShareToken(sessionId);
return {
url: `${process.env.NEXT_PUBLIC_VERCEL_PROJECT_PRODUCTION_URL}/share/${shareToken}`,
token: shareToken,
};
}
Sandbox Lifecycle
// packages/sandbox/src/lifecycle.ts
export class SandboxLifecycle {
async create() {
this.sandbox = await createSandbox({
baseSnapshotId: this.baseSnapshotId,
});
return this.sandbox;
}
async hibernate() {
const snapshot = await this.sandbox.createSnapshot();
await this.sandbox.stop();
return snapshot.id;
}
async resume(snapshotId: string) {
this.sandbox = await createSandbox({
baseSnapshotId: snapshotId,
});
return this.sandbox;
}
async destroy() {
await this.sandbox.stop();
this.sandbox = null;
}
}
Troubleshooting
Workflow Execution Fails
Problem: Workflow steps timeout or fail
// Check workflow logs
const logs = await getWorkflowLogs(workflowId);
console.log(logs);
// Increase timeout
export const POST = createWorkflow(
async (context, data) => {
// ...
},
{ timeout: 300000 } // 5 minutes
);
Sandbox Connection Issues
Problem: Cannot connect to sandbox
// Verify sandbox is running
const status = await sandbox.status();
console.log('Sandbox status:', status);
// Recreate sandbox
await sandbox.stop();
const newSandbox = await createSandbox();
GitHub App Permissions
Problem: Cannot access repositories
- Verify app installation:
https://github.com/settings/installations - Check app permissions include:
- Repository contents: Read & write
- Pull requests: Read & write
- Metadata: Read-only
Database Migrations
Problem: Schema out of sync
# Check migration status
bun run ci
# Apply migrations manually if needed
cd apps/web
npx drizzle-kit push:pg
Authentication Issues
Problem: OAuth callback fails
- Verify callback URLs match exactly (including protocol)
- Check
BETTER_AUTH_SECRETis set and consistent across deployments - Confirm OAuth app credentials are correct
Port Conflicts in Sandbox
Problem: Preview port already in use
// Use alternative ports
const ports = [3000, 5173, 4321, 8000];
for (const port of ports) {
try {
await sandbox.exec(`lsof -ti:${port} | xargs kill -9`);
} catch {
// Port was free
}
}
Development Commands
# Start development server
bun run web
# Lint and format check
bun run check
# Auto-fix linting and formatting
bun run fix
# Type check all packages
bun run typecheck
# Full CI suite
bun run ci
# Refresh sandbox base snapshot
bun run sandbox:snapshot-base
API Reference
Core Types
// Agent context passed to tools
interface AgentContext {
sandbox: Sandbox;
sessionId: string;
userId: string;
metadata: Record<string, unknown>;
}
// Tool definition
interface Tool<T = unknown> {
name: string;
description: string;
parameters: z.ZodSchema<T>;
execute: (params: T, context: AgentContext) => Promise<unknown>;
}
// Sandbox interface
interface Sandbox {
exec(command: string): Promise<ExecResult>;
readFile(path: string): Promise<string>;
writeFile(path: string, content: string): Promise<void>;
search(pattern: string): Promise<SearchResult[]>;
createSnapshot(): Promise<Snapshot>;
status(): Promise<SandboxStatus>;
stop(): Promise<void>;
}
Resources
- Homepage: https://open-agents.dev
- Repository: https://github.com/vercel-labs/open-agents
- Vercel Workflow SDK: https://vercel.com/docs/workflow
- Better Auth: https://www.better-auth.com/
- GitHub Apps: https://docs.github.com/en/apps
FAQ
What does the vercel-open-agents skill do?
Build and run durable background coding agents with workflow orchestration, isolated sandboxes, and GitHub integration on Vercel.
How do I install vercel-open-agents?
Run npx skills add https://github.com/reason-machines/ai-agent-skills --skill vercel-open-agents in your project. The skills CLI detects the agents you use and copies the skill into the right folder. For Claude Code you can also place the skill folder at ~/.claude/skills/vercel-open-agents/ (personal) or .claude/skills/vercel-open-agents/ (project).
Which AI agents can use vercel-open-agents?
Any agent that supports the open Agent Skills format (a folder with a SKILL.md file), including Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Windsurf, Cline, OpenCode and more.
Is vercel-open-agents safe to install?
No partner security audit is available yet. Skills can run commands and scripts, so read the SKILL.md and any bundled files before installing.
Who maintains vercel-open-agents?
It is published by reason-machines in reason-machines/ai-agent-skills. It has 165 recorded installs.
More from reason-machines/ai-agent-skills
agency-agents-zh-skill
reason-machines/ai-agent-skills
Open the skill to read its SKILL.md instructions.
agent-reach-internet-access
reason-machines/ai-agent-skills
Open the skill to read its SKILL.md instructions.
agent-sprite-forge-2d-game-assets
reason-machines/ai-agent-skills
Generate 2D sprite sheets, animated GIFs, transparent PNGs, layered maps, and engine-ready prototypes for Godot/Unity using Codex-driven asset pipelines.
agentmemory-persistent-memory
reason-machines/ai-agent-skills
Add persistent memory to AI coding agents using agentmemory - remembers context, preferences, and decisions across sessions
awesome-agent-skills
reason-machines/ai-agent-skills
Open the skill to read its SKILL.md instructions.
oh-my-openagent-orchestration
reason-machines/ai-agent-skills
Open the skill to read its SKILL.md instructions.
Popular devops & cloud skills
microsoft-foundry
microsoft/azure-skills
Build, deploy, evaluate, optimize, fine-tune, and manage Microsoft Foundry agents, models, and resources end to end. USE FOR: foundry, azd ai agent, azd provision/deploy, hosted agent scaffold/develop/run/deploy/troubleshoot, prompt agent create, managed harness agent create, GitHub Copilot harness agent, create agent, update agent, add tool to agent, invoke agent, agent insights, evaluate agent, batch eval, continuous eval, continuous monitoring, agent CI/CD, optimize prompt, improve prompt, optimize agent instructions, Agent Optimizer scaffold, dataset curation from traces, deploy model, model fine-tuning (SFT/DPO/RFT), Foundry project, RBAC, role assignment, permissions, quota, capacity, region, deployment failure, AI Services, create Foundry resource, knowledge index, customize deployment, onboard, availability, training-data, grader, distillation, large file upload. DO NOT USE FOR: Azure Functions, App Service, general Azure deploy (use azure-deploy), general Azure prep (use azure
azure-diagnostics
microsoft/azure-skills
Debug Azure production issues on Azure using AppLens, Azure Monitor, resource health, and safe triage. WHEN: debug production issues, troubleshoot app service, app service high CPU, app service deployment failure, troubleshoot container apps, container app 503 or target port mismatch, troubleshoot functions, function telemetry missing or stopped after a config/connection-string change, troubleshoot AKS, VM RDP, Linux SSH, VM black screen, can't connect to VM, reset VM password, NSG or firewall blocking, kubectl cannot connect, kube-system/CoreDNS failures, pod pending, crashloop, node not ready, upgrade failures, analyze logs, KQL, insights, image pull failures, cold start issues, health probe failures, resource health, root cause of errors, troubleshoot event hubs, troubleshoot service bus, messaging SDK error, AMQP connection failure, message lock lost, lock renewal during long-running processing, service bus dead letter.
azure-prepare
microsoft/azure-skills
Prepare azd-based Azure projects for deployment: generates azure.yaml, infrastructure (Bicep/Terraform), and Dockerfiles for the Azure Developer CLI (azd) workflow. USE ONLY when the user explicitly wants to use azd as the deployment tool, or the project already has an azure.yaml file. DO NOT USE FOR: non-azd deployments, Python App Service code-only deploys (use python-appservice-deploy), or cross-cloud migration (use azure-cloud-migrate). WHEN: prepare app for azd, create azure.yaml, set up azd infrastructure, modernize app for Azure with azd, deploy with azd, function app, timer trigger, service bus trigger, event-driven function, managed identity, generate Bicep, generate Terraform, create and deploy to Azure.
azure-ai
microsoft/azure-skills
Use for Azure AI: Search, Speech, OpenAI, Document Intelligence. Helps with search, vector/hybrid search, speech-to-text, text-to-speech, transcription, OCR. WHEN: AI Search, query search, vector search, hybrid search, semantic search, speech-to-text, text-to-speech, transcribe, OCR, convert text to speech.
azure-deploy
microsoft/azure-skills
Execute Azure deployments for ALREADY-PREPARED applications that have existing .azure/deployment-plan.md and infrastructure files. DO NOT use this skill when the user asks to CREATE a new application — use azure-prepare instead. This skill runs azd up, azd deploy, terraform apply, and az deployment commands with built-in error recovery. Requires .azure/deployment-plan.md from azure-prepare and validated status from azure-validate. WHEN: "run azd up", "run azd deploy", "execute deployment", "push to production", "push to cloud", "go live", "ship it", "bicep deploy", "terraform apply", "publish to Azure", "launch on Azure". DO NOT USE WHEN: "create and deploy", "build and deploy", "create a new app", "set up infrastructure", "create and deploy to Azure using Terraform" — use azure-prepare for these.
azure-validate
microsoft/azure-skills
Pre-deployment validation for Azure readiness. Run deep checks on configuration, infrastructure (Bicep or Terraform), RBAC role assignments, managed identity permissions, and prerequisites before deploying. WHEN: validate my app, check deployment readiness, run preflight checks, verify configuration, check if ready to deploy, validate azure.yaml, validate Bicep, test before deploying, troubleshoot deployment errors, validate Azure Functions, validate function app, validate serverless deployment, verify RBAC roles, check role assignments, review managed identity permissions, what-if analysis, validate Container Apps deployment.