Now in Closed Beta

Run AI Coding Agents Securely. Build Workflows on Top.

Connect your repo, call our API, get results. We handle isolation, security, and cost limits — you ship your product.

Sign Up for Beta
Scroll

Run from your favourite place

One API call triggers a secure agent run. Plug it into whatever you already use.

Zapier

Zapier

Connect agent runs to 6 000+ apps

Make

Make

Add agent runs to any workflow

n8n

n8n

Trigger runs from self-hosted workflows

Github

GitHub Actions

Trigger on push, PR, or schedule

REST API

Call from any language or tool

ModelContextProtocol

MCP

Use from Claude, Cursor, or Windsurf

Built for security. Designed for builders.

Everything you need to run AI coding agents safely in production.

One API Call

One endpoint handles provisioning, cloning, execution, and cleanup. You just send a prompt.

Terminal
curl -X POST https://api.cospec.io/v1/runs \
  -H "Authorization: Bearer $COSPEC_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "template": "node",
    "repo": "acme/app",
    "prompt": "Fix failing tests and create a PR",
    "model": "sonnet",
    "guardrails": {
      "maxCostUsd": 1
    }
  }'

Sandbox Isolation

Every run gets its own sandbox. Nothing shared, nothing persisted. Agents can't escape.

  • Secure by default
  • Deterministic environment
  • No cleanup, no drift

Guardrails

Set limits before every run. Stay in control.

maxCostUsd1.00
maxTurns50
maxDurationSec300

Templates

Pre-configured environments with custom env vars. Pick a stack, start coding.

nodeNode 22 + npm
pythonPython 3.12 + pip

Full Audit Trail

Every action — logged and queryable.

12:01:03CLONEacme/app
12:01:07EXEC npm install
12:01:14WRITEsrc/auth.ts
12:01:18NET api.github.com

Network Control

Allowlist the exact domains agents can reach.

github.comallow
npmjs.orgallow
*.unknown.iodeny

Frequently Asked Questions

Everything you need to know before we start

What exactly does coSPEC do?

coSPEC provides secure execution infrastructure for AI coding agents. You connect a repo, send a prompt, and build whatever you need on top — we take care of sandboxing, network control, and guardrails.

How is this different from Devin or GitHub Copilot Agents?

Those are end-user products — you use them as-is. coSPEC is infrastructure you build on. We provide the secure execution engine; you build your own workflows, triggers, and UX on top. Your product, your rules.

Why not just use a raw VM or sandbox?

Raw sandbox providers give you a VM and nothing else. You'd still need to build the agent execution layer, security controls, cost limits, and audit trails yourself. coSPEC gives you all of that out of the box, purpose-built for code agents.

What can I build with coSPEC?

Anything that involves running AI coding agents: auto-fix CI failures, automated PR reviews, code migration tools, internal developer productivity platforms. If it involves an AI agent analysing, writing, or modifying code, coSPEC can run it securely.

How do I integrate coSPEC?

API-first. A single API call starts a run. We also offer Zapier, Make, and n8n nodes, a GitHub Action, and an MCP server — so you can plug coSPEC into whatever workflow you already use.

How much does coSPEC cost?

Every beta user gets $25 in free credits to start. We'll introduce pricing as the product matures — beta users will get early access to plans and a say in how pricing is shaped.

What does the beta include?

Full API access with every feature included — no gated tiers. You get direct access to the founding team and a voice in shaping the product. We're building this with our beta users, not just for them.

Still have questions?

Sign Up for BetaContact Us