PanDev Metrics Review

Four IDE plugins, seven integrations, DORA and SPACE scoring, on-prem via Docker or Kubernetes. That's a real architecture, not a marketing wrapper. The free PanDev CLI — local-only, no telemetry, sha256-verifiable — is the kind of detail that signals engineers built this for engineers.

Two things worry me. No pricing page, no changelog, no API. "Contact sales" on the main product tells me this is early, maybe pre-revenue-model, not necessarily broken but unproven.

The obvious alternative here is self-reported status updates and sprint velocity guesses. PanDev replaces that with event-level data, which is a real advance, not just a cheaper dashboard. But IDE telemetry on individual developers is a culture landmine if rolled out without buy-in. Pilot with one team, watch the trust reaction before you touch org-wide metrics.

Event-level capture from VS Code, JetBrains, Cursor, and Visual Studio plugins is the right architecture for engineering intelligence — self-reported status updates are the thing every DORA rollout dies on, and IDE-level telemetry sidesteps that. Pairing it with GitHub, GitLab, Bitbucket, Azure DevOps, Jira, ClickUp, and Linear integrations means the metrics layer isn't guessing at delivery data from a single source.

The on-prem option via Docker or Kubernetes is the line item that matters for fintech and govtech. If I'm running a regulated shop, I don't ship IDE keystroke-adjacent data to a third-party SaaS without that escape hatch. Pricing for the team platform is contact-sales-only, though, which is friction I have to plan around before I can even scope a pilot.

The free PanDev CLI — no account, no network calls, sha256-verifiable — is a smart trust-building wedge, but it's a separate tool solving a narrower problem: AI agent spend, not org-wide delivery metrics. Adopt this and in three years you've got IDE-level observability baked into how you manage engineers, for better or worse with developer trust.

$0 for the CLI. Real number, verifiable — tcpdump and sha256 checksums, not marketing claims. That part's honest.

The platform tier is labeled "Free" but requires "contacting sales or booking a demo." That's not free. That's undisclosed. No seat count, no per-user rate, no on-prem premium listed anywhere in the evidence. Can't build a 3-year TCO model on contact-us. Team of 50 could land at $15K or $150K — no way to know before the call.

On-prem via Docker or Kubernetes adds deployment and maintenance cost the SaaS tier skips — evidence doesn't say how much, but self-hosting never runs free labor-wise. Compare that to the CLI's zero-friction install (npx pandev@beta). Same company, two very different procurement paths.

Paging at 3am, I care about MTTR and change failure rate being accurate, not vibes-based. PanDev pulls those from IDE events plus Git/Jira/Linear rather than asking engineers to self-report, which is the right instinct — self-reported status data rots fast under incident pressure. On-prem via Docker or Kubernetes matters too; fintech/govtech shops that can't ship IDE telemetry off-network get a real option, not a promise.

Day-3 concern: IDE plugins across four editors is real surface area to keep patched and low-overhead. Rolling out Cursor/JetBrains/VS Code agents without perf complaints from devs is nontrivial, and the evidence doesn't show telemetry footprint numbers.

The CLI is the sleeper feature: no account, no network calls, tcpdump-verifiable, reflog-based cost attribution at ~90% accuracy. That's the kind of tool an SRE trusts on-call. Tradeoff: pricing is 'contact sales' for the actual platform, so I can't size cost against DORA-tooling alternatives yet.

The free CLI piece is the clever bit here. `npx pandev@beta` reads your Claude Code or Codex logs locally, no account, no network calls, and you can literally run tcpdump to check they're not lying about it. That's a rare bit of honesty in a category full of vague telemetry promises. Nice touch letting you verify a sha256 checksum yourself.

The bigger platform, though, the one with IDE plugins for VS Code, JetBrains, Cursor, and Visual Studio, is listed as 'Free [POPULAR]' but pricing is actually 'contact sales.' No pricing page, no free trial. That's homework before you even get to try it, and the site has no changelog either, so you can't tell how fast this thing moves.

DORA and SPACE scoring from real IDE events beats self-reported status reports, no argument. The tradeoff: no visible track record on reliability, and mobile isn't mentioned anywhere in the evidence, which for a monitoring tool matters less day-to-day but still stings if you want a phone check on deploy metrics.

Started with the CLI, because that's the part I can actually check. No account, no network calls, sha256 checksum via `pandev privacy` against reproducible-build hashes. That's a real privacy claim, not a marketing one. Good sign.

The platform side is a different story. DORA metrics, SPACE scores, DevEx indicators, IDE plugins across four editors, seven integrations — broad claims, thin proof. No pricing page. 'Teams / Enterprise' is labeled Free but requires a sales call, which usually means it isn't. No changelog listed either, so I can't tell how fast this ships.

Engineering intelligence has a graveyard — self-reported status tools that engineers gamed or ignored. IDE-level telemetry is the honest fix, if it holds up. On-prem via Docker/Kubernetes is a real exit lever for regulated shops. Everyone else's dashboards die with the vendor; yours might not, if you self-host from day one.