AgentTeams multi-agent teams.

UltraCode-style multi-agent orchestration: a generatable, savable, governable, observable, resumable workflow layer.

Skill-driven harness/loop engineering workflow agent plugin.

Scheduled coding runs in fresh agent sessions with auditable history.

Shared multi-agent task board (create/claim/transition/query) over a Cordis service key.

Adaptive deep-research orchestrator built on the official workflow engine.

Claude Science-style research workbench: ReAct research-loop engine (research_* tools), versioned artifacts with provenance (artifact_* tools), and 10 science skills for genomics/pathogens/bioinformatics.

Pair a second model that passively reviews each turn and injects notes.

Condition-driven wakeup: durable file/command/http/process/webhook watches that wake the agent.

Per-project scheduled prompts run as fresh headless agent sessions, with one-time, interval, and cron schedules plus durable run history.

Model-based permission approval: an approval-request answerer backed by a separate reviewer model.

Embedded task management engine: decision-point protocol, idea capture wall, Linear-style issue store.

Reusable agent squads with per-agent provider/model routes and tool policies, serial/parallel dispatch, spawn/fork/chain context modes, and a Web management panel.

Global persistent agent squads with per-agent model/tool policies: manage them in Settings, select and toggle one per conversation, then collaborate on normal sends in a fixed or model-planned order.

Research agent suite: long-horizon goal control with auditable evidence chains, cron scheduling, multi-agent expert teams, self-evolving research memory (FTS5 + RRF recall), project workspaces, and a custom workspace UI.

Adversarial checkup → fix → review loop toolset.

Plan review with anchored annotations and structured feedback back to the agent.

Recurring loops: `/loop` command + loop tool + activity status bar.

Adds an `auto` permission preset between workspace-write and danger-full-access: a classifier grants routine sandbox escalations once, while dangerous or uncertain requests still go to a human.

Second-model auto-review on the approval answerer chain: a read-only reviewer subagent returns structured allow/deny verdicts with reasons, fail-closed by default.

Unified anchored agent preset for DeepSeek V4 Pro/Flash on the official API and opencode-go: Minimal-exact two-tool first-request bootstrap, model-aware Pro/Flash paths, and epoch-aware catalog management for long-session stability.

Adds an Auto Approve permission preset to the Web UI, using a fresh restricted Reviewer Agent to allow or deny each approval request.

Task planning with experience muscle-memory: condition-reflex recall of past solutions, LLM capability matching, and auto-persisted lessons.

Two-phase agent preset for DeepSeek V4 Pro: the first request sees only bash+read (Minimal-aligned bootstrap), then the full Standard tool catalog after the first tool call or reply; ships as an installer bundle plus a manual preset directory.