Forty-plus engines across six product pillars. Observability, control, memory, orchestration, extensibility, and team analytics — all over the same trace data, all working on any AI coding tool.
The foundation. Everything your AI coding tool is doing, visible the moment it happens — no scrollback hunting, no black boxes, no after-the-fact forensics.
Real-time token gauge, per-file relevance scores, and drop alerts. See what the AI can still see — and re-inject any dropped file with one click.
Real-time agent tree for every spawned process, tool execution, and sub-agent task. Pause, resume, or kill directly from the UI.
AI-decomposed step breakdown from a single goal. Steps auto-complete from trace matching. Divergence detection flags plan drift.
Detects three failure patterns in real time: same file edited 5+ times, 8+ consecutive searches without edits, 4+ tool failures in a row. Severity escalates on repeated triggers.
Daemon-backed PTY survives app restarts, view switches, and reboots. Snapshot-based attach restores any historical session state instantly.
Per-model pricing, live cost gauge, and an explicit confidence badge that distinguishes API-reported usage from heuristic estimates.
Interactive playback with scrubber, speed control (1x/2x/4x/8x), and code/conversation toggle. Share any session with a link — redaction controls included.
Trace-grouped timeline with category filters, sort toggles, and a properties panel for deep trace inspection.
Stuck developer detection, rabbit hole alerts, budget spikes, context collapse warnings, and repeat tool failures — surfaced the moment patterns emerge.
Seeing isn’t enough. Define what the agent can touch, review every plan step before it continues, and rollback to any checkpoint safely. The agent works inside your rules, at OS level.
If the agent goes off-limits, FSWatcher catches the write in ~100ms. Operon reverts the file, pauses the PTY via SIGTSTP, and shows you the diff with approve or reject.
Between plan steps, review all changes before the agent proceeds. Named git snapshots at each gate let you rollback to any checkpoint with one click.
Continuous comparison between declared intent and observed activity from EventPipeline tool events. Heuristic checks (out-of-scope ratio, file count, consecutive errors) fire the moment they diverge.
File count, dependency depth, scope proximity, context pressure, historical success rate, and code complexity — scored and surfaced at every gate before you approve.
Auto-detects patterns in checkpoint rejections and scope violations. After N rejections in a directory, suggests scope rules to prevent repeat mistakes.
Checkpoint manager creates git tags at each gate. Approve, reject, or rollback to any prior snapshot — full rollback without losing work on other sessions.
Decisions, patterns, failure modes, peak hours — built from your own sessions, not a generic model. The instruments adapt to how you actually work.
Auto-extracted from conversations via Claude Haiku. Searchable, tagged, and scored by confidence. Never lose the reasoning behind a choice — even across sessions and team members.
Chronological memory view with superseded values. Toggle between grid and timeline. Understand what the project remembered, and when it changed its mind.
Jaccard-similarity detection surfaces memory conflicts in real time. Keep-existing or use-new resolution with a full audit trail of every merge decision.
At session end, Operon writes AGENTS.md to the repo with captured decisions, patterns, and project context. If the file is Operon-generated, it updates in place.
Peak hours, tool preferences, plan abandonment, failure modes, context exhaustion rate — built from your sessions. The more you fly, the smarter the instruments get.
Complexity assessment, risk signals, codebase alerts, and suggested steps — debounced as you type your goal. A briefing that prevents problems instead of reacting to them.
Seven-rule analysis (file refs, action verbs, length, outcome, scope, pronouns, testing) runs on your goal text. Badge + inline suggestions in the New Session dialog.
Matches the current step against traces from similar past steps. Returns success probability, likely failure modes, and estimated time. Wired into every checkpoint review.
Pure SQL aggregation over your session history. Per-directory difficulty map, approach fingerprinting (plan-first / test-first / iterate / explore-first / greenfield), and cost prediction from keyword-similar past sessions.
Enter a goal. Operon decomposes into subtasks, recommends tools per task type, dispatches work, and handles the handoff — including worktrees and remote machines.
Heuristic goal → subtask DAG based on keyword patterns (add / fix / refactor / migrate / test / document). Tool recommendations per subtask.
Scores each tool by task-type keyword matching and your historical success rate from the traces table. Claude Code for architecture, Cursor for UI, Aider for quick fixes — from your own data.
Full plan lifecycle: create, start, pause, resume, cancel. Session linking, onSessionEnded advances subtasks, plan cards with progress bars in the Orchestration view.
In-repo, new-worktree, pooled (warm pool), or remote (SSH). Pool warmth indicator in the New Session dialog. Default-preservation cleanup and reclaim-on-archive — worktrees don’t leak.
RemoteSessionBridge drops remote shell output into the same EventBus as local. Remote git (clone / fetch / worktree add / commit) over SSH. Connection pool with DB-backed credentials + injectable ssh2 factory.
When session A (tool X) ends and session B (tool Y) is active in the same project, Operon surfaces a context bridge panel with cross-tool info. Auto-dismisses after 60s, dismissed IDs persisted.
One adapter system for every AI tool. One MCP hub for every agent. Skills, automations, webhooks, and issue imports — so Operon slots into the stack you already have.
Claude Code (hooks), Cursor / Codex / Gemini CLI / Aider (PTY regex + stream JSON). AdapterRegistry auto-detects installed binaries. EventPipeline feeds all tracks into one Trace model.
Install MCP servers from one place; Operon writes the right config into every agent profile (Claude Code / Cursor / Gemini / Aider). First-write confirmation dialog before touching any config file.
Parse skill manifests, install into agent profiles, bind per-agent. Full RPC CRUD + Settings UI. Browse, install, and remove without leaving Operon.
AutomationEngine runs on cron schedules or trigger events. Pluggable ActionRegistry — custom action adapters hang off the registry. Per-automation status, last run, dry-run preview.
WebhookReceiver verifies HMAC signatures on every inbound webhook. Failed signature = 401, no silent drops. Trigger automations from PR events, issue updates, CI failures.
Shared issue schema across three providers. IssueImporter fans out to all sources with per-source failure isolation. IssuePicker wired into the New Session dialog — seed a session with full issue context.
Heuristic body generator: files grouped by section, decisions → rationale block, TODOs → follow-ups list. Copy-to-clipboard or (roadmap) push to GitHub via the integration client.
Cost dashboards, best practice extraction, knowledge sharing, tool benchmarking, diff quality, code impact — team-wide intelligence over the same trace data.
Four dashboards: per-member cost with weekly trends, best practice extraction (members with ≥70% success rate per tool+category), knowledge sharing from hard directories, and per-tool benchmarking across the team.
Per-file snapshots at session end, cross-session churn detection over a 30-day lookback via git diffNameOnly. Detects files AI edited that were later manually modified.
Static analysis with 4 heuristic checks (large functions >50 lines, explicit any types, TODO/FIXME comments, commented-out code). Deduction scoring, wired into every completed trace.
Per-session, per-project, per-developer attribution. Model-tier breakdown (Opus / Sonnet / Haiku). Budget alerts with warning, critical, and exceeded thresholds.
Cost intelligence, productivity velocity, code impact heatmaps, session quality KPIs, tool usage distribution, smart alerts — via recharts, scoped to session or project.
Multi-tenant project membership with role scoping. PM dashboard with real-time team activity. One-click client reports — sessions, decisions, deliverables.
Filterable by actor, resource, and time range. Shipped with the Superadmin web app — search, suspend, impersonate, feature-flag toggles.
13 syncable tables with outbox-pattern push/pull, conflict resolution, and Supabase Realtime for instant team sync. RLS enforced on every cloud_* table.
Operon doesn’t pick winners. All instruments above work identically regardless of which tool you use. The adapter system routes data from any CLI into the same Trace model — via hooks where supported, PTY regex everywhere else.
Full hooks API — structured tool names, inputs, outputs, sub-agents, session context. Richest data path.
PTY interception with regex patterns. Works with Cursor’s terminal without any config.
PTY + stream JSON parsing for structured output with —output-format stream-json.
PTY output regex analysis. ANSI-cleaned, idle-timeout trace completion.
PTY + git diff integration. Auto-detects binary, health-checked on session start.
Want to see the full adapter + integrations story? See /integrations.
Private beta — join the waitlist to get an invite when your slot opens up.