Group: GitHub JuneLeGency Harness

Sort by:Popular▼

✓Popular✓Newest

In-app agent harness: a model plus composable capabilities (tools, skills, MCP, sub-agents) as one ChatBackend, mirroring Pydantic AI Harness.

Shell commands for in-app agents (run_command, start_command, check_command, stop_command) over a pluggable runtime, mirroring Pydantic AI Harness Shell.

Web browsing for in-app agents on an off-screen WebView (navigate, snapshot, click, type_text, press_key, select_option, hover, wait_for, get_text, scroll, go_back, go_forward), mirroring the Pydantic AI Harness browser tools.

Device capabilities for in-app agents: device info, clipboard, calendar, contacts, location, alarms and timers, notifications, with runtime permissions and approvals.

Sandboxed workspace file tools for in-app agents (read_file, write_file, edit_file, list_directory, search_files, find_files, create_directory, file_info), mirroring Pydantic AI Harness FileSystem.

Persistent agent memory (MEMORY.md notebook with write_memory, read_memory, delete_memory, search_memory), mirroring Pydantic AI Harness Memory.

Model-owned task plans (write_plan, read_plan, add_task, update_task_status(es), remove_task) with a live Plan element, mirroring Pydantic AI Harness Planning.

An Alpine Linux sandbox for in-app agents: PRoot (GPL-2.0, bundled as a separate executable) running an Alpine minirootfs, as a ShellRuntime for harness-shell.

Scheduled and recurring agent runs for in-app agents (schedule_task, list_scheduled_tasks, cancel_scheduled_task) on WorkManager.

Text to speech for in-app agents (speak, stop_speaking) on the platform TextToSpeech engine.