01
Bring your agent
Today that is Pi Agent — add an LLM key in Settings. Describe the job: refactor, migrate, report, or preview.
About
Your agent. Our cloud.
ExeCode hosts coding agents in an isolated cloud sandbox. Available now: Pi Agent with your API key — multi-step work online with files, shell, previews, and a clear record of what ran. Sandboxes are session-based and time-limited, not always-on. Roadmap: Codex, Claude Code, and Grok Build using your own subscription, plus OpenCode and more. Pro includes 100 Run tasks per month (included models — Grok 4.5, GPT-5.6, and more) and 50 hosted Agent sessions (your LLM key).
Example jobs:Refactor + patch notesMigrate and planStripe MRR reportPublish dashboard to here.now
How it works
Your agent. Our cloud.
Pi Agent is live with your API key. Next up: Codex, Claude Code, and Grok Build using your own subscription — plus OpenCode and other agent runtimes.
02
We run it in the cloud
ExeCode hosts an isolated cloud sandbox: the agent writes and runs code, uses the shell, and iterates across files. Jobs are session-based and time-limited — not always-on. More agents on the roadmap.
03
Get work and a record back
Download files, charts, previews, or hosted links — plus a clear record of what ran.
Capabilities
Built for people who want outcomes.
Pro is live chat with real execution: included Run models (Grok 4.5, GPT-5.6, and more), Python and Node in isolated sandboxes, file uploads, generated artifacts, connected APIs, and clear monthly limits on Run and Agent.
Connect real tools
Add your own Stripe, OpenRouter, Airtable, or here.now key, then ask for a report or publish in chat.
Plain English in
Describe the outcome you want. ExeCode writes the script, calls the API, and runs it for you.
Artifacts out
Get processed output and downloadable CSVs, markdown reports, charts, files, or hosted links back in chat.
Capability packs
Playbooks for known-good jobs — reports, comparisons, dashboards, and exports — when you ask in chat.
Iterates when things fail
If an API call, script, or build fails, ExeCode inspects the error, patches the code, and reruns — especially powerful in Agent mode for multi-file projects.
Safe & isolated
Execution happens in secure Vercel Sandbox microVMs with your saved secrets injected only for the task.
Connected APIs
Use your keys for reports, comparisons, and publishes.
Add a key in Settings → Secrets, then ask in chat. ExeCode runs that job in the sandbox and returns files — not a scheduled or always-on workflow. Agent LLM keys go in AI Providers.
Stripe revenue reports
STRIPE_SECRET_KEY
Ask: “Generate my MRR and revenue report for the last 12 months.”
Get: CSV export, markdown summary, customer revenue tables, and charts.
For indie hackers and SaaS founders who want answers from Stripe without writing billing scripts.
OpenRouter model cost reports
OPENROUTER_API_KEY
Ask: “Compare these models for cost, speed, and output quality.”
Get: Markdown report, cost table, latency notes, and a practical recommendation.
For AI builders choosing models without getting mugged by token pricing.
here.now publishing
HERENOW_API_KEY
Ask: “Build a tiny dashboard and publish it.”
Get: Static page, hosted URL, source files, and generated artifacts.
For solo builders who need a shareable result, not a deployment ceremony.
Two surfaces
Chat for quick tasks. Work for agent projects.
Both surfaces return real output — not just suggestions. Chat uses included Run models (Grok 4.5 default). Work is the Pi agent workspace with projects, threads, live files, and preview — BYOK at Settings → AI Providers.
Chat
Quick tasks, your choice of model
ExeCode writes a script, runs it on a secure sandbox, and shows the result. Choose from included models — Grok 4.5 (default), GPT-5.6 Sol, Terra, and Luna, and more. No LLM API key required. Counts toward 100 runs/month.
- Stripe revenue reports
- Data cleanup and charts
- File extraction and conversion
- Model cost comparisons
Work
Pi agent workspace — projects, threads, isolated sandbox
Multi-step Pi Agent jobs in an isolated cloud sandbox. Send @pi with your LLM key (OpenRouter, OpenAI, Anthropic, xAI, or Google). Each project has its own files; each @pi job is a thread. Sandboxes are session-based and time-limited — not always-on. 50 new sessions/month; follow-ups in the same active session are free.
- Build and test multi-file projects
- Debug and fix errors in the thread
- Research, then build, then verify
- Iterate until the result is right
Switch Chat | Work anytime from the sidebar.
Pricing
One subscription. Two ways to execute.
Pro includes included Run models for quick tasks and hosted Agent sandboxes for multi-step work — with honest monthly limits.
ExeCode Pro
$29/month
100 runs + 50 Agent sessions per billing period. Real output in chat — not just code.
ExeCode Pro
$29/month
50 hosted Agent sessions + 100 Run tasks — compute and sandbox, not tokens
- Run mode with included models — Grok 4.5, GPT-5.6, and more (switch in chat)
- 100 Run tasks per month — scripts, projects, and live previews
- Agent mode — your LLM key + 50 hosted sandbox sessions per month
- Use your Stripe, OpenRouter, Airtable, and here.now keys in chat
- Chat history, file uploads, artifacts, and capability pack playbooks
Agent sessions
50 hosted Agent sandbox sessions per month
- Pi Agent live today — add your LLM key in Settings → AI Providers
- One session = one new isolated cloud sandbox for multi-step work
- Follow-up messages in the same active Agent chat do not use another session
- Friendly reminder at 40 sessions. New Agent sandboxes pause at 50 until the next billing period
Included Run models
Grok 4.5, GPT-5.6, and more — no API key needed
- Default model: Grok 4.5
- GPT-5.6 Sol, Terra, and Luna (OpenAI)
- Six GPT-5.6 variants on OpenRouter (standard + Pro tiers)
- Switch models from the Run mode menu in chat
- Model access is included with Pro — separate from your run quota
Included runs
100 task runs per month included
- Each time ExeCode runs a script or builds a project, that's one run
- Included Run models and sandbox compute are part of each run
- Runs reset at the start of each billing period
- Friendly reminder at 80 runs. New runs pause at 100 until the next billing period
Connected APIs (BYOK)
Your LLM key for Agent; API keys for connected services
- API keys (Stripe, Airtable, etc.) in Settings → Secrets
- Agent LLM keys in Settings → AI Providers
- Run mode uses included server models — no LLM API key required
- In Agent mode, add an OpenRouter key (or OpenAI, Anthropic, xAI, Google) and pick your model in chat
- You pay those providers directly — ExeCode does not mark up API usage
- Mention Stripe, Airtable, or similar in chat after you add the key
- 50 hosted Agent sandbox sessions per billing period (your LLM key)
- Pi Agent live today — Codex, Claude Code, Grok Build on the roadmap
- 100 secure Run tasks with included models (Grok 4.5, GPT-5.6, and more)
- Friendly warnings at 80 runs and 40 Agent sessions
- AI provider fallback when the Sandbox queue is busy
- Cancel anytime via the Polar customer portal
FAQ
Quick answers.
What is ExeCode?
Your agent, our cloud. Bring your own LLM key; ExeCode hosts an isolated cloud sandbox where a coding agent can write and run code, use the shell, and return files, previews, and a clear record of what ran — not just suggestions. Run mode also supports quick one-shot jobs with included models. Sandboxes are per session and time-limited; they do not run unattended overnight.
Which agents can I run?
Available now: Pi Agent with your own API key (Settings → AI Providers). On the roadmap: Codex, Claude Code, and Grok Build using your own subscription, plus OpenCode and more agent runtimes. ExeCode hosts an isolated, time-limited cloud sandbox; you keep the AI relationship.
What models can I use in Run mode?
Pro includes server-hosted models you pick in chat: Grok 4.5 (default), GPT-5.6 Sol, Terra, and Luna, plus six GPT-5.6 variants via OpenRouter. You do not add an LLM API key for Run — model access is included with your subscription.
What counts as one run?
Every time ExeCode runs code or builds a project in Run mode, that counts toward your 100/month allowance. Plain messages, reading files, and web searches do not count.
What counts as one Agent session?
Each time ExeCode starts a new Agent sandbox VM, that counts toward your 50/month allowance. Follow-up messages in the same active session are free until the sandbox expires (about 15 minutes on the current Vercel path). Sandboxes stop when idle — they do not stay running overnight. You bring your own LLM API key.
Do I need API keys?
Run mode includes Grok 4.5, GPT-5.6, and more — no LLM key required. Add API keys (Stripe, Airtable, etc.) in Settings → Secrets when you use connected services. Agent mode requires at least one LLM key in Settings → AI Providers.
What's the difference between Run and Agent?
Run is fast, one-shot execution with included models (Grok 4.5, GPT-5.6, and more). Agent is multi-step: plan, write, test, fix, and verify across files. Agent today is Pi with your model key; ExeCode hosts an isolated, time-limited sandbox. More agent runtimes are on the roadmap.
Do I pay ExeCode for OpenRouter or Stripe API usage?
No. You pay third-party providers directly. ExeCode charges one flat Pro subscription for chat, sandbox hosting, and orchestration. OpenRouter models in Run mode are included with Pro when enabled on the deployment.
Do I need to install anything?
No. All execution runs in an isolated cloud sandbox — subscribe, connect your keys, and work from chat.
Ready to run your agent in the cloud?
Subscribe to Pro — 100 runs and 50 Agent sessions per month. Bring your own key for Agent; we host an isolated cloud sandbox and return files, previews, and a clear record of what ran.
Already subscribed? Open chat and put your agent to work.