Environments and sandbox providers

Early access. Cloud sessions and managed agent runs are not open to every account yet: we are letting accounts in gradually while we test. They run in a cloud sandbox operated by Mutagent, so there is nothing to host. A sandbox with nothing to do for 15 minutes stops; send an interactive session a message and it wakes up, delivers your message and carries on in the same conversation. A -p task run cannot be woken; run it again.

An Environment holds the variables and secrets your cloud sessions’ tools need, such as a GitHub token or a database URL. The Python SDK lists, describes, creates, updates, replaces and deletes Environments, and lists the sandbox providers and presets a Helix Cloud launch can name. An Environment has variables and secrets only. It has no tools: tools come with the agent, see Tools and skills. Model keys do not go in an Environment; they come from your LLM providers. Environment calls are on client.environments. The token exchange, sandbox provider and preset calls are on client.sandbox. Request bodies and responses are typed models from mutagent.models; read response fields as attributes. These pages require mutagent-sdk 0.4.1.

Get a sandbox token

Environment and sandbox provider calls take a short-lived sandbox token, not your API key. Exchange the API key for a token with create_sandbox_token. Pass the token as bearer_token to a second client. A client takes either api_key or bearer_token, not both.

from mutagent import Mutagent
from mutagent.models import SandboxTokenRequest

SERVER_URL = "https://api.mutagent.io"

platform = Mutagent(server_url=SERVER_URL)  # reads MUTAGENT_API_KEY
exchange = platform.sandbox.create_sandbox_token(SandboxTokenRequest(workspace_id="<workspace-id>"))

client = Mutagent(server_url=SERVER_URL, bearer_token=exchange.token)
print(f"token valid for {exchange.expires_in} seconds")

Response field	What it is
`token`	The sandbox token.
`expires_in`	Seconds until the token expires. Exchange the API key again after that. Do not store the token.

workspace_id is the workspace’s ID; platform.workspaces.list_workspaces() returns it as workspaces[].id_. The API key must be allowed to create resources in that workspace, and a key limited to one workspace or organization can only get a token inside it. Refusals: 401, 403, 422, 429.

List Environments

list_environments returns every Environment in the workspace.

from mutagent import Mutagent
from mutagent.models import SandboxTokenRequest

SERVER_URL = "https://api.mutagent.io"
token = Mutagent(server_url=SERVER_URL).sandbox.create_sandbox_token(
    SandboxTokenRequest(workspace_id="<workspace-id>")
).token
client = Mutagent(server_url=SERVER_URL, bearer_token=token)

for environment in client.environments.list_environments().environments:
    print(environment.name, len(environment.vars_), len(environment.secrets))

Each Environment has this shape. No call returns a stored value.

Field	What it is
`name`	The Environment name, such as `ci` or `prod`.
`vars_`	The variables, each with `name` and `fingerprint`.
`secrets`	The secrets, each with `name` and `fingerprint`. Secrets are stored encrypted.
`created_at`, `updated_at`	ISO timestamps, or `None`.

A fingerprint is 8 hex characters the server computes from the value with a key only the server holds. Two equal fingerprints mean the same value; a changed fingerprint means the value changed. You cannot compute a fingerprint yourself, so hashing your local value will not match it. To be sure a stored value is right, set it again.

Describe one Environment

get_environment returns one Environment in the same shape.

from mutagent import Mutagent
from mutagent.models import SandboxTokenRequest

SERVER_URL = "https://api.mutagent.io"
token = Mutagent(server_url=SERVER_URL).sandbox.create_sandbox_token(
    SandboxTokenRequest(workspace_id="<workspace-id>")
).token
client = Mutagent(server_url=SERVER_URL, bearer_token=token)

ci = client.environments.get_environment("ci")
print([f"{entry.name} {entry.fingerprint}" for entry in ci.vars_])

404 when the workspace has no Environment with that name.

Create an Environment or set entries

update_environment creates the Environment when it does not exist, and otherwise changes only the entries you name. A string value sets an entry. None removes it. Entries you do not name are kept.

from mutagent import Mutagent
from mutagent.models import EnvironmentPatchRequest, SandboxTokenRequest

SERVER_URL = "https://api.mutagent.io"
token = Mutagent(server_url=SERVER_URL).sandbox.create_sandbox_token(
    SandboxTokenRequest(workspace_id="<workspace-id>")
).token
client = Mutagent(server_url=SERVER_URL, bearer_token=token)

ci = client.environments.update_environment(
    "ci",
    EnvironmentPatchRequest(
        vars_={"DATABASE_URL": "<database-url>", "OLD_FLAG": None},
        secrets={"GITHUB_TOKEN": "<github-token>"},
    ),
)
print(ci.name, ci.updated_at)

It returns the Environment as get_environment does.

Replace an Environment

replace_environment sets the whole Environment. Entries the body does not name are deleted, secrets included.

from mutagent import Mutagent
from mutagent.models import EnvironmentReplaceRequest, SandboxTokenRequest

SERVER_URL = "https://api.mutagent.io"
token = Mutagent(server_url=SERVER_URL).sandbox.create_sandbox_token(
    SandboxTokenRequest(workspace_id="<workspace-id>")
).token
client = Mutagent(server_url=SERVER_URL, bearer_token=token)

prod = client.environments.replace_environment(
    "prod",
    EnvironmentReplaceRequest(
        vars_={"DATABASE_URL": "<database-url>"},
        secrets={"GITHUB_TOKEN": "<github-token>"},
    ),
)
print(len(prod.vars_), len(prod.secrets))

Rules for both calls

Rule	Refusal
The Environment name matches `^[A-Za-z0-9._-]{1,64}$`.	422
A variable or secret name matches `^[A-Z_][A-Z0-9_]*$`. A value is a string, or `None` to remove the entry when you set entries.	422
One Environment holds at most 64 KiB.	422
A variable named like an LLM provider key, such as `ANTHROPIC_API_KEY`, is refused unless you pass `allow_provider_key=QueryFlag.VALUE_1`, the value `"1"`. `QueryFlag` is in `mutagent.models`. With it, the value replaces the workspace key for runs that load this Environment.	409

Delete an Environment

delete_environment deletes the Environment and its secrets.

from mutagent import Mutagent
from mutagent.models import SandboxTokenRequest

SERVER_URL = "https://api.mutagent.io"
token = Mutagent(server_url=SERVER_URL).sandbox.create_sandbox_token(
    SandboxTokenRequest(workspace_id="<workspace-id>")
).token
client = Mutagent(server_url=SERVER_URL, bearer_token=token)

result = client.environments.delete_environment("ci")  # returns a dict
print(result["deleted"], result["name"])

It returns {"deleted": True, "name": ...}. 404 when there is no Environment with that name, including a delete you retry. Sessions already running keep the values they started with. A later launch that names the deleted Environment is refused with 404.

List sandbox providers

list_sandbox_providers lists the sandbox providers a launch can name, and the default.

from mutagent import Mutagent
from mutagent.models import SandboxTokenRequest

SERVER_URL = "https://api.mutagent.io"
token = Mutagent(server_url=SERVER_URL).sandbox.create_sandbox_token(
    SandboxTokenRequest(workspace_id="<workspace-id>")
).token
client = Mutagent(server_url=SERVER_URL, bearer_token=token)

sandbox_providers = client.sandbox.list_sandbox_providers()
print(sandbox_providers.names, sandbox_providers.default)

Response field	What it is
`names`	The sandbox provider names.
`default`	The sandbox provider a launch uses when it names none, or `None`.

Pass a name as sandbox_provider when you launch a session. The SDK has no call that adds or configures a sandbox provider. See Sandbox providers.

List presets

list_sandbox_presets lists the presets a launch can name as preset, and the default.

from mutagent import Mutagent
from mutagent.models import SandboxTokenRequest

SERVER_URL = "https://api.mutagent.io"
token = Mutagent(server_url=SERVER_URL).sandbox.create_sandbox_token(
    SandboxTokenRequest(workspace_id="<workspace-id>")
).token
client = Mutagent(server_url=SERVER_URL, bearer_token=token)

listing = client.sandbox.list_sandbox_presets()
for preset in listing.presets:
    print(preset.name, preset.arch, preset.provider, preset.environment)
print(listing.default)

Field	What it is
`presets[].name`	The preset name.
`presets[].description`	What the preset runs.
`presets[].arch`	The CPU architecture.
`presets[].provider`	The sandbox provider the preset runs on, when it names one.
`presets[].environment`	The Environment the preset loads, when it names one.
`default`	The preset a launch uses when it names none, or `None`.

Errors

A refused call raises SDKError. status_code is the HTTP status and body is the JSON the server returned, as text, with error and message.

import json

from mutagent import Mutagent, SDKError
from mutagent.models import SandboxTokenRequest

SERVER_URL = "https://api.mutagent.io"
token = Mutagent(server_url=SERVER_URL).sandbox.create_sandbox_token(
    SandboxTokenRequest(workspace_id="<workspace-id>")
).token
client = Mutagent(server_url=SERVER_URL, bearer_token=token)

try:
    client.environments.get_environment("missing")
except SDKError as err:
    print(err.status_code, json.loads(err.body)["message"])

Async

AsyncMutagent has the same calls on client.environments and client.sandbox; await each one. The same Environments from the command line: mutagent env.

​Get a sandbox token

​List Environments

​Describe one Environment

​Create an Environment or set entries

​Replace an Environment

​Rules for both calls

​Delete an Environment

​List sandbox providers

​List presets

​Errors

​Async

Get a sandbox token

List Environments

Describe one Environment

Create an Environment or set entries

Replace an Environment

Rules for both calls

Delete an Environment

List sandbox providers

List presets

Errors

Async