{
  "markdown": "# AgentTasker MCP Server\n\n<!-- mcp-name: io.github.S3bRR/agent-tasker-mcp -->\n\nAgentTasker is a small, stdio-only MCP server for AI agents that need to run multiple tasks quickly and get structured results back in one call.\n\nIt is intentionally narrow:\n\n- two tools: `execute` and `execute_batch`\n- local stdio transport only\n- zero third-party runtime dependencies\n- explicit dependency control with `depends_on`\n- compact, model-friendly JSON responses\n\nRepository: `https://github.com/S3bRR/agent-tasker-mcp`\n\n## Why This Exists\n\nMost agent orchestration layers are heavier than they need to be. This project is designed for the common case:\n\n- run a few tasks in parallel\n- let one task wait on another when needed\n- keep the MCP surface small enough for models to use reliably\n\nThere is no queue service, no persistence layer, no background worker system, and no SDK dependency required at runtime.\n\n## What It Supports\n\nTask types:\n\n- `python_code`\n- `http_request`\n- `discovery_search`\n- `web_scrape`\n- `shell_command`\n- `file_read`\n- `file_write`\n\nPublic MCP tools:\n\n- `execute`\n- `execute_batch`\n\n## Install\n\nRequirements:\n\n- Python 3.10+\n- A local MCP client that can run stdio servers\n\n### Recommended: `uvx`\n\nRun directly from GitHub:\n\n```bash\nuvx --from git+https://github.com/S3bRR/agent-tasker-mcp.git agent-tasker-mcp-server --workers 8\n```\n\nOnce the package is live on PyPI, the command becomes:\n\n```bash\nuvx agent-tasker-mcp-server --workers 8\n```\n\n### `pipx`\n\nInstall directly from GitHub:\n\n```bash\npipx install git+https://github.com/S3bRR/agent-tasker-mcp.git\n```\n\nOnce the package is live on PyPI, the command becomes:\n\n```bash\npipx install agent-tasker-mcp-server\n```\n\n### Local clone\n\n```bash\ngit clone https://github.com/S3bRR/agent-tasker-mcp.git\ncd agent-tasker-mcp\n./setup.sh\n```\n\n`setup.sh` creates a local `.venv`, installs this package into it, and prints an\nabsolute MCP config snippet. If `python3 -m venv` is not available, it falls back\nto `virtualenv` when installed.\n\n## MCP Client Configuration\n\n### GitHub Source\n\n```json\n{\n  \"command\": \"uvx\",\n  \"args\": [\n    \"--from\",\n    \"git+https://github.com/S3bRR/agent-tasker-mcp.git\",\n    \"agent-tasker-mcp-server\",\n    \"--workers\",\n    \"8\"\n  ]\n}\n```\n\n### Installed Package\n\n```json\n{\n  \"command\": \"agent-tasker-mcp-server\",\n  \"args\": [\"--workers\", \"8\"]\n}\n```\n\n### Local checkout\n\n```json\n{\n  \"command\": \"/absolute/path/to/agent-tasker-mcp/.venv/bin/agent-tasker-mcp-server\",\n  \"args\": [\"--workers\", \"8\"]\n}\n```\n\nUse the exact absolute path printed by `./setup.sh` for local checkouts.\n\n## Usage\n\n### `execute`\n\nRun one task immediately.\n\n```json\n{\n  \"task_type\": \"python_code\",\n  \"code\": \"result = 6 * 7\"\n}\n```\n\n### `execute_batch`\n\nRun multiple tasks concurrently.\n\n```json\n{\n  \"tasks\": [\n    {\n      \"name\": \"fetch_users\",\n      \"task_type\": \"http_request\",\n      \"url\": \"https://api.example.com/users\"\n    },\n    {\n      \"name\": \"calc\",\n      \"task_type\": \"python_code\",\n      \"code\": \"result = 6 * 7\"\n    }\n  ],\n  \"output_mode\": \"compact\"\n}\n```\n\n### `depends_on`\n\nIf one task must wait for another, make it explicit.\n\n```json\n{\n  \"tasks\": [\n    {\n      \"name\": \"write_file\",\n      \"task_type\": \"file_write\",\n      \"path\": \"/tmp/example.txt\",\n      \"content\": \"hello\"\n    },\n    {\n      \"name\": \"read_file\",\n      \"task_type\": \"file_read\",\n      \"path\": \"/tmp/example.txt\",\n      \"depends_on\": [\"write_file\"]\n    }\n  ]\n}\n```\n\nIf an upstream dependency fails, downstream tasks are marked failed and do not run.\n\n## Output Shape\n\n`output_mode` supports:\n\n- `compact` (default)\n- `full`\n\nThe response is ordered to match the input task list, which makes it easier for models to consume without extra reconciliation logic.\n\n## Release Process\n\nReleases are tag-driven.\n\n1. update `pyproject.toml` and `server.json` to the same version\n2. commit and push to `main`\n3. create and push a matching tag such as `v1.0.0`\n4. GitHub Actions runs tests, builds the package, publishes to PyPI through Trusted Publishing, and then publishes `server.json` to the MCP Registry\n\nThe release workflow rejects version drift: the pushed tag, `pyproject.toml`, and `server.json` must match exactly.\n\n## Limits\n\nOptional environment variables:\n\n- `AGENT_TASKER_MAX_TASKS`: maximum tasks per `execute_batch`\n- `AGENT_TASKER_MAX_PAYLOAD_BYTES`: maximum payload size per task\n- `AGENT_TASKER_MAX_MEMORY_MB`: soft process memory guard\n\n## Security Notes\n\nThis server is intended for trusted environments.\n\n- `python_code` executes Python code\n- `shell_command` executes shell commands\n- `file_read` and `file_write` operate on the local filesystem\n\nDo not expose this server directly to untrusted users.\n\n## Development\n\nCreate a local environment:\n\n```bash\n./setup.sh\nsource .venv/bin/activate\n```\n\nRun the server:\n\n```bash\nagent-tasker-mcp-server --workers 4\n```\n\nRun tests:\n\n```bash\n.venv/bin/python -m unittest discover -s tests\n```\n\n## Packaging\n\nThis repo includes [server.json](./server.json) for MCP Registry publication and a GitHub Actions workflow that publishes both the PyPI package and MCP metadata from a version tag.\n\n## License\n\nMIT\n",
  "bytes": 5107,
  "sha": "bc75aa63aa7aecf15958e5cf8934225fd3f73ec205f3f0b0514a383db1271b46",
  "repo_slug": "s3brr/agent-tasker-mcp",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/mcp_io_github_s3brr_agent_tasker_mcp_041bb1f8/readme"
}