name: skill-as-a-service description: | Run coding agents (Claude Code, Gemini, Codex) with skills in the cloud via API. Each task gets its own isolated VM. Use when you need to spawn sub-tasks, delegate work to other agents, or run tasks with specific skills. Requires REBYTE_API_KEY environment variable.

Skill-as-a-Service API

Spawn coding agent tasks in the cloud via API. Each task gets its own isolated VM with skills pre-installed.

Before You Start

Check that REBYTE_API_KEY is set:

echo "$REBYTE_API_KEY"

If empty, ask the user for their API key (get one at https://app.rebyte.ai/settings/api-keys), then:

export REBYTE_API_KEY="rbk_..."

Create a Task

curl -s -X POST https://api.rebyte.ai/v1/tasks \
  -H "API_KEY: $REBYTE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "Your task description here",
    "skills": ["deep-research"]
  }'

Response:

{
  "id": "550e8400-...",
  "workspaceId": "660e8400-...",
  "url": "https://app.rebyte.ai/run/550e8400-...",
  "status": "running",
  "createdAt": "2026-02-09T10:30:00.000Z"
}

The url is live immediately. Give it to the user — they watch the agent work in real time. You do NOT need to poll or wait.

Create a Task with Files

If the agent needs to work on files (PDFs, CSVs, images, etc.):

# 1. Get a signed upload URL
FILE_RESP=$(curl -s -X POST https://api.rebyte.ai/v1/files \
  -H "API_KEY: $REBYTE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"filename": "data.csv"}')

UPLOAD_URL=$(echo "$FILE_RESP" | jq -r '.uploadUrl')
FILE_ID=$(echo "$FILE_RESP" | jq -r '.id')
FILE_NAME=$(echo "$FILE_RESP" | jq -r '.filename')

# 2. Upload the file content
curl -s -X PUT "$UPLOAD_URL" \
  -H "Content-Type: application/octet-stream" \
  --data-binary @data.csv

# 3. Create a task with the file attached
curl -s -X POST https://api.rebyte.ai/v1/tasks \
  -H "API_KEY: $REBYTE_API_KEY" \
  -H "Content-Type: application/json" \
  -d "{
    \"prompt\": \"Analyze this CSV and summarize findings\",
    \"skills\": [\"data-analysis\"],
    \"files\": [{\"id\": \"$FILE_ID\", \"filename\": \"$FILE_NAME\"}]
  }"

The file is copied into the task's VM at /code/{filename} before the agent starts.

Create a Task with a GitHub Repo

curl -s -X POST https://api.rebyte.ai/v1/tasks \
  -H "API_KEY: $REBYTE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "Add unit tests for the auth module",
    "skills": ["deep-research"],
    "githubUrl": "owner/repo",
    "branchName": "main"
  }'

Share Results Publicly

By default, tasks are visible only to org members. To share with anyone (no login required):

TASK_ID="the-task-id-from-create"
curl -s -X PATCH "https://api.rebyte.ai/v1/tasks/$TASK_ID/visibility" \
  -H "API_KEY: $REBYTE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"visibility": "public"}'

Response includes shareUrl — give this to anyone:

{
  "visibility": "public",
  "shareUrl": "https://app.rebyte.ai/share/550e8400-..."
}

Check Task Status (Optional)

You don't need to poll — the URL is live. But if you need to wait for completion:

curl -s "https://api.rebyte.ai/v1/tasks/$TASK_ID" \
  -H "API_KEY: $REBYTE_API_KEY" | jq '{status, url}'

Statuses: running, completed, failed, canceled

Follow Up on a Task (Optional)

Send additional instructions to a running or completed task:

curl -s -X POST "https://api.rebyte.ai/v1/tasks/$TASK_ID/prompts" \
  -H "API_KEY: $REBYTE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Now fix the issues you found"}'

Reuse a Workspace (Optional)

Pass workspaceId from a previous task to skip VM provisioning (much faster):

# First task creates a new VM
TASK1=$(curl -s -X POST https://api.rebyte.ai/v1/tasks \
  -H "API_KEY: $REBYTE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Set up the project", "githubUrl": "owner/repo"}')
WS_ID=$(echo "$TASK1" | jq -r '.workspaceId')

# Second task reuses the same VM
curl -s -X POST https://api.rebyte.ai/v1/tasks \
  -H "API_KEY: $REBYTE_API_KEY" \
  -H "Content-Type: application/json" \
  -d "{\"prompt\": \"Now add tests\", \"workspaceId\": \"$WS_ID\"}"

Delete a Task

curl -s -X DELETE "https://api.rebyte.ai/v1/tasks/$TASK_ID" \
  -H "API_KEY: $REBYTE_API_KEY"

Returns HTTP 204 (no content).

Create Task Parameters

| Field | Type | Required | Description | |-------|------|----------|-------------| | prompt | string | Yes | Task description (max 100,000 chars) | | executor | string | No | opencode (default), claude, gemini, codex | | model | string | No | Model tier (default: lite) | | files | object[] | No | Files from POST /v1/files. Each: {"id": "...", "filename": "..."} | | skills | string[] | No | Skill slugs: ["deep-research", "pdf", "data-analysis"] | | githubUrl | string | No | GitHub repo (owner/repo) | | branchName | string | No | Branch (default: main) | | workspaceId | string | No | Reuse a workspace from a previous task |

All Endpoints

| Method | Endpoint | Description | |--------|----------|-------------| | POST | /v1/files | Get signed upload URL for a file | | POST | /v1/tasks | Create a task | | GET | /v1/tasks | List tasks | | GET | /v1/tasks/:id | Get task with status and prompts | | POST | /v1/tasks/:id/prompts | Send a follow-up prompt | | PATCH | /v1/tasks/:id/visibility | Set private/shared/public | | DELETE | /v1/tasks/:id | Delete task |

Bundled Python Client

This skill includes a Python client and CLI at scripts/. To use them:

# Find the skill directory
SKILL_DIR=$(find ~/.skills -maxdepth 1 -name '*skill-as-a-service*' -type d | head -1)

# Use the CLI
python3 "$SKILL_DIR/scripts/rebyte_cli.py" create --prompt "Hello world"
python3 "$SKILL_DIR/scripts/rebyte_cli.py" get TASK_ID
python3 "$SKILL_DIR/scripts/rebyte_cli.py" list

# Or use the client in Python
python3 -c "
import sys; sys.path.insert(0, '$SKILL_DIR/scripts')
from rebyte_client import RebyteClient
client = RebyteClient()
task = client.create_task(prompt='Hello world')
print(task['url'])
"

See references/api.md for full API details and references/examples.md for more examples.