capsule-run 0.8.1

Secure WASM runtime to isolate and manage AI agent tasks

capsule-run

A secure, durable runtime for agentic workflows

Overview

Capsule is a runtime for coordinating AI agent tasks in isolated environments. It is designed to handle untrusted code execution, long-running workflows, large-scale processing, or even multi-agent systems.

Each task runs inside its own WebAssembly sandbox, providing:

• Isolated execution: Each task runs isolated from your host system
• Resource limits: Set CPU, memory, and timeout limits per task
• Automatic retries: Handle failures without manual intervention
• Lifecycle tracking: Monitor which tasks are running, completed, or failed

This enables safe task-level execution of untrusted code within AI agent systems.

Installation

pip install capsule-run

Getting started

Create hello.py:

from capsule import task

@task(name="main", compute="LOW", ram="64MB")
def main() -> str:
    return "Hello from Capsule!"

Run it:

capsule run hello.py

Use --verbose to display real-time task execution details.

Production

Running source code directly (like .py) evaluates and compiles your file at runtime. While great for development, this compilation step adds a few seconds of latency. For use cases where sub-second latency is critical, you should build your tasks ahead of time.

# Generates an optimized hello.wasm file
capsule build hello.py --export

# Execute the compiled artifact directly
capsule exec hello.wasm

[!NOTE] Or from your existing code:

from capsule import run

result = await run(
   file="./hello.wasm", # or `hello.py`
   args=[]
)

print(f"Task completed: {result['result']}")

See Integrate Into an Existing Project for details.

Executing a .wasm file bypasses the compiler completely, reducing initialization time to milliseconds while using a natively optimized (.cwasm) format behind the scenes.

Integrate Into an Existing Project

The run() function lets you execute tasks programmatically from your application code, no CLI needed.

from capsule import run

result = await run(
    file="./capsule.py", # or `capsule.wasm`
    args=["code to execute"]
)

Create capsule.py:

from capsule import task

@task(name="main", compute="LOW", ram="64MB")
def main(code: str) -> str:
    return exec(code)

How It Works

Simply annotate your Python functions with the @task decorator:

from capsule import task

@task(name="analyze_data", compute="MEDIUM", ram="512MB", timeout="30s", max_retries=1)
def analyze_data(dataset: list) -> dict:
    """Process data in an isolated, resource-controlled environment."""
    return {"processed": len(dataset), "status": "complete"}

The runtime requires a task named "main" as the entry point. Python will create one automatically if none is defined, but it's recommended to set it explicitly.

When you run capsule run main.py, your code is compiled into a WebAssembly module and executed in a dedicated sandbox.

Response Format

Every task returns a structured JSON envelope containing both the result and execution metadata:

{
  "success": true,
  "result": { "processed": 5, "status": "complete" },
  "error": null,
  "execution": {
    "task_name": "data_processor",
    "duration_ms": 1523,
    "retries": 0,
    "fuel_consumed": 45000,
    "ram_used": 1200000,
    "host_requests": [{...}]
  }
}

Response fields:

• success — Boolean indicating whether the task completed successfully
• result — The actual return value from your task (json, string, null on failure etc.)
• error — Error details if the task failed ({ error_type: string, message: string })
• execution — Performance metrics:
  • task_name — Name of the executed task
  • duration_ms — Execution time in milliseconds
  • retries — Number of retry attempts that occurred
  • fuel_consumed — CPU resources used (see Compute Levels)
  • ram_used — Peak memory used in bytes
  • host_requests — List of host requests made by the task

Documentation

Task Configuration Options

Parameter	Description	Type	Default	Example
name	Task identifier	str	function name	"process_data"
compute	CPU level: "LOW", "MEDIUM", "HIGH"	str	"MEDIUM"	"HIGH"
ram	Memory limit	str	unlimited	"512MB", "2GB"
timeout	Maximum execution time	str	unlimited	"30s", "5m"
max_retries	Retry attempts on failure	int	0	3
allowed_files	Folders accessible in the sandbox (with optional access mode)	list	[]	["./data"], [{"path": "./data", "mode": "ro"}]
allowed_hosts	Domains accessible in the sandbox	list	[]	["api.openai.com", "*.anthropic.com"]
env_variables	Environment variables accessible in the sandbox	list	[]	["API_KEY"]

Compute Levels

• LOW: Minimal allocation for lightweight tasks
• MEDIUM: Balanced resources for typical workloads
• HIGH: Maximum fuel for compute-intensive operations
• CUSTOM: Specify exact fuel value (e.g., compute="1000000")

Project Configuration (Optional)

Create a capsule.toml file in your project root to set default options:

[workflow]
name = "My AI Workflow"
version = "1.0.0"
entrypoint = "src/main.py"  # Run `capsule run` without specifying a file

[tasks]
default_compute = "MEDIUM"
default_ram = "256MB"
default_timeout = "30s"

Task-level options always override these defaults.

File Access

Tasks can read and write files within directories specified in allowed_files. Any attempt to access files outside these directories is not possible.

allowed_files supports directory paths only, not individual files.

Each entry can be a plain path (read-write by default) or a dict with an explicit mode: "read-only" (or "ro") or "read-write" (or "rw").

from capsule import task

@task(name="main", allowed_files=[
    {"path": "./data", "mode": "read-only"},
    {"path": "./output", "mode": "read-write"},
])
def main() -> str:
    with open("./data/input.txt") as f:
        content = f.read()
    with open("./output/result.txt", "w") as f:
        f.write(content)
    return content

Plain strings are still accepted: allowed_files=["./output"] defaults to read-write.

Network Access

Tasks can make HTTP requests to domains specified in allowed_hosts. By default, no outbound requests are allowed ([]). Provide an allowlist of domains to grant access, or use ["*"] to allow all domains.

Wildcards are supported: *.example.com matches all subdomains of example.com.

from capsule import task
from urllib.request import urlopen
import json

@task(name="main", allowed_hosts=["api.openai.com", "*.anthropic.com"])
def main() -> dict:
    with urlopen("https://api.openai.com/v1/models") as response:
        return json.loads(response.read().decode("utf-8"))

Environment Variables

Tasks can access environment variables to read configuration, API keys, or other runtime settings. Use Python's standard os.environ to access environment variables:

from capsule import task
import os

@task(name="main", env_variables=["API_KEY"])
def main() -> dict:
    api_key = os.environ.get("API_KEY")
    return {"api_key": api_key}

Compatibility

✅ Supported:

• Pure Python packages and standard library
• json, math, re, datetime, collections, etc.

⚠️ Not yet supported (inside the sandbox):

• Packages using C extensions require a wasm32-wasi compiled wheel (e.g. numpy, pandas)

These limitations only apply to the task file executed in the sandbox. Your host code using run() has access to the full Python ecosystem, any pip package, native extensions, everything. (see Integrate Into an Existing Project)

Links

• GitHub
• Issues