Kova

The OpenClaw runtime validation lab.

Kova proves OpenClaw works end-to-end on real machines: install, upgrade, gateway, sessions, plugins, agent turns, providers, dashboards, TUIs, MCP, browsers, soak — with real execution evidence, not synthetic confidence.

Can real users install, update, start, message, use plugins, and keep running without OpenClaw getting slow, unhealthy, leaky, or broken?

What Kova Proves

🚀 Install & upgrade	Fresh installs, release-track upgrades, version-to-version, durable-env clone → local-build, migrations.
🧩 Every runtime path	Gateway, sessions, plugins, agent (CLI + Gateway), dashboard, TUI, MCP, browser, OpenAI-compatible, soak.
💥 Failure containment	Timeouts, malformed providers, streaming stalls, network offline, missing auth, concurrent load, recovery.
⏱ Honest agent latency	Pre-provider · provider · post-provider, split per turn with timeline span evidence.
🧠 Process attribution	CPU and RSS across 21 named roles — never just "memory was high".
📈 Baselines & regressions	--repeat N with median / p95 / max / variance. Per-platform baselines. Gate refuses to ship without proof.
🔍 Inventory audit	Discovers OpenClaw CLI, scripts, plugins, entrypoints, and repeated Kova run work.
🛰 Diagnostics contract	Timeline spans, event-loop delay, CPU/heap profiles on --deep-profile.
🤖 Agent-first I/O	kova.report.v1 JSON, summary, bundle, paste, compare. Verdict-led dashboard by default.

42 scenarios   ·   29 surfaces   ·   26 states   ·   8 profiles
21 process roles   ·   15+ collectors

Quickstart

npm install
node bin/kova.mjs setup
node bin/kova.mjs self-check

Run a release gate against a local OpenClaw build:

node bin/kova.mjs matrix run \
  --profile release \
  --target local-build:/path/to/openclaw \
  --execute --gate

Kova data lives in ~/.kova (credentials, reports, artifacts, baselines).

The Commands You'll Use

kova plan                       # what Kova will do, by profile/scenario/state
kova inventory plan             # find OpenClaw capabilities Kova doesn't cover yet
kova inventory repeated-work    # find duplicated scenario commands and collector pressure
kova run --scenario <id>        # one scenario, one target
kova matrix run --profile <p>   # the release matrix (multi-target, --repeat N)
kova reports                    # recent reports and short run IDs
kova report <runId|run.json>    # the dashboard above
kova report compare <a> <b>     # baseline vs current with regression deltas
kova report bundle <run.json>   # portable evidence pack for handoff
kova report paste <run.json>    # fixer-ready prompt
kova help <command>             # per-command detail

Every command renders a dashboard by default. Add --json for machines, --plain for compact text, --no-progress to silence streaming, or --ascii for Unicode-free output. Color, width, NO_COLOR, and CI runners are auto-detected.

Targets

npm:<version>              published OpenClaw release
release:<name>             published release track (stable, beta, …)
runtime:<name>             existing OCM runtime
local-build:<repo-path>    local checkout built as a release-shaped runtime

Profiles

smoke                  fast confidence over core paths
diagnostic             local-build with timeline + span expectations
release                ship / no-ship gate coverage
soak                   long-running pressure and stability
release-upgrade        published release-track upgrade matrix
local-build-upgrade    upgrade into a local build
official-plugins       bundled + official plugin coverage
exhaustive             the full sweep (--allow-exhaustive)

Safety

Dry-run by default. Real execution requires --execute. Disposable envs are destroyed by default. Durable user envs are clone sources, never mutation targets. Exhaustive profiles require --allow-exhaustive. Keep a failing lab with --retain-on-failure when you need to look.

For Agents

kova plan --json
kova inventory plan --openclaw-bin openclaw --openclaw-repo /path/to/openclaw --json
kova inventory repeated-work --json
kova matrix plan --profile smoke  --target runtime:stable --json
kova matrix run  --profile smoke  --target runtime:stable --execute --json

inventory plan --openclaw-repo includes a source/catalog drift check for the OpenClaw message-channel capability catalog. Runtime scenarios do not read the OpenClaw source tree; they probe the selected release-shaped package.

Repo-local agent skills ship in .agents/skills/:

• kova-operator — benchmark workflows, evidence rules, report handoff.
• ocm-operator — safe env cloning, local runtime builds, service inspection.

Learn More

• What Kova is — the model, in one page.
• Scenario hierarchy — surfaces, states, scenarios.
• Diagnostics contract — what OpenClaw emits and how Kova uses it.
• Report schema — the kova.report.v1 JSON contract.
• Agent usage — the agent-first workflow.

Kova uses OCM as the harness. OpenClaw is the product under test.