code_session
codesession.exec / codesession.artifacts — Execute stateful Python sessions and persist session files/artifacts
How to use it
code_session is exposed by the Arena Agent MCP server. Add the server to your MCP client (Claude Desktop, Cursor, Windsurf and others), and the code_session tool becomes available to the model automatically. See the full listing for setup details and every tool this server provides.
FULL ARENA AGENT LISTINGOther tools in Arena Agent (39)
admin.run — Cross-platform admin escalation. Linux/macOS proxies to sudo
asr.transcribe — Transcribe an audio file locally with whisper.cpp. Auto-conv
browser.search — DuckDuckGo search via pure-Python (no chromium)
capabilitygap.record / capabilitygap.list / capabilitygap.resolve — Track missing bridge capabilities found during real scenarios
code.run — Execute agent-authored code under the operator's execution posture (composable fence); fail-closed, agent can't set the posture
codeartifact.read — Read a persisted Code Workbench artifact by runid and path
codematrix.run — Run up to 8 Code Workbench jobs sequentially under the current operator posture
codeproject.run / codeproject.lockverify / codeproject.promotetool — Run persistent projects, enforce dependency locks, or promote recipes/tests into tools
coderun.info / coderun.promotetool — Inspect persisted runs or use a run as provenance for a promoted tool
desktop.ocr — Run OCR on a fresh desktop screenshot and return recognized
desktopapp.clickwindowrelative / desktopapp.screenshotwindow — Resolve a real window, then click or screenshot relative to that window for less brittle GUI scenarios
document.structure — Structure OCR/ASR/text into tasks or physics homework JSON
emulator.providers / emulator.list / emulator.start / emulator.stop / emulator.attach — Boot and stop Android emulators through whichever manager the host has (AVD, Genymotion, MuMu, Waydroid, or a host-declared one); post-boot control is plain ADB via mobile.. See docs/emulators.md
exec.exec — Namespaced alias for exec. Run shell command outside bri
fs.read — Read file contents (utf-8)
git.status — Show git status for a repository.
hooks.list — List configured hooks per event
image.preprocessforocr — Preprocess an image file for OCR
inputhelper.click / inputhelper.key / inputhelper.launch / inputhelper.sendchatcommand — Route real hardware input through the Interactive Input Helper running in the user's desktop session (solves Session 0 limitation for Java Swing, LWJGL, and all GUI automation)
mcp.extcall — Call a tool on a registered external MCP server (Desktop-Commander, ScreenPilot, ...). Discover servers with mcp.extservers and their tools with mcp.exttools
mcpserver.create / mcpserver.test / mcpserver.install — Author, verify, and install an external MCP stdio server
memory.recall — Find relevant facts/snapshots/sessions by query (TF score),
mission.autopilotstart / mission.autopilotreport / mission.run — Execute bounded mission tool chains with persisted progress and flight records, or run persisted missions
mobile.preflight / mobile.devices — Preflight Android/ADB readiness and list connected devices
net.http — Typed HTTP client. Only http/https to public hostnames (inhe
ocr.extract — Run OCR on any image file and return text + word boxes
plan.create — Create a structured execution plan for a goal, with suggeste
react.run — Run a bounded reason-act-observe loop using safe observation
reflect.run — Reflect on a prior react/planning run and produce concerns,
relay.check / relay.reply / relay.send — Operator mailbox: read instructions left from the Dashboard or a terminal, answer them, or ask a question mid-task. Queued, not pushed — the agent has to look.
runtime.probe / runtime.compat — Probe runtimes and show runtime × sandbox compatibility with known blockers and next actions
scenario.run / scenario.promotefromhistory — Execute scenarios or promote successful runs/history into reusable scenarios
secrets.list — List available secret keys (values never returned).
service.autostartstatus / service.autostartrepair — Diagnose or repair bridge autostart setup
skill.list — List available agent skills
subagent.spawn — Spawn isolated subagent for delegated work; returns summary
sudo.run — Run a command through 'sudo -n <cmd>' (non-interactive). Req
sys.status — Bridge/services/funnel status
watch.files — List, add, or remove file watchers that emit realtime file c