All software should be self-improving.

Polylane reads what your software does in production and sends the fix back as a pull request.

Writing code is cheap now.

Coding agents ship more in a day than your team can read in a week.

Nobody can keep up with production anymore.

Hundreds of services across clouds nobody remembers setting up, changing all day. It's too big to hold in your head and it moves faster than you can read.

The hard part starts after you ship.

Agents made writing code fast. Deploys, alerts, incidents and regressions still land on the same few people.

Polylane closes the loop.

It reads what production is telling you and sends the change back as a pull request. Every release makes the next one better.

What Polylane does after you ship.

Knows your whole system.

Every service, database and queue across your clouds, and what depends on what.

The context graph→
orders-db · What depends on it
orders-db PlanetScale database Down
orders-api Kubernetes deployment direct
web Vercel project 2 hops
hd-prod Hyperdrive config direct
checkout-edge Cloudflare Worker 2 hops
billing-export Lambda function direct

Ignores the noise.

Every alert is checked against what's normal for that service. False alarms get dismissed before anyone is paged.

p99 latency over threshold on search
Nightly reindex, back to normal by 02:20
Triaging Dismissed
TimeoutError in POST /checkout
checkout-edge · 283 events
Triaging Confirmed
CPUUtilization high on batch-worker
Scheduled backfill
Triaging Dismissed

Finds what you didn't think to monitor.

Polylane reads your code and watches checkouts, sign-ups and payments in your real traffic.

src/payments/stripe.ts
41 const charge = await stripe.charges.create(params);
42 if (charge.status === "failed") {
43 metrics.increment("payment.failed", {
44 reason: charge.failure_code,
45 });
46 }
New signal: payment.failed per minute
Watches failures on the payments route, from line 43

Proves the root cause.

Agents investigate in parallel and check each other's work before you see an answer.

How swarms work→
Deploy 9f3c2a1 cut the hd-prod pool from 50 to 5
Agent 2 · checked by agent 1
Proposed Confirmed
The 502s come from the edge
Agent 3 · checked by agent 2
Proposed Ruled out
Traffic spiked at 20:14
Agent 4 · checked by agent 2
Proposed Ruled out

Writes the fix.

You get a pull request with what broke and why. If CI fails, Polylane fixes that too.

How fixes work→
Pull request
Open coreplanelabs/checkout-edge #1318
fix(checkout-edge): restore the Hyperdrive pool size
What broke

Checkout p99 rose 18x at 20:14. Requests were queueing for a database connection.

Why

Deploy 9f3c2a1 lowered max_connections on hd-prod from 50 to 5.

The fix

Restores the pool to 50 and adds a test that fails if it drops below peak concurrency.

Polylane review 3 rounds, nothing open
CI 18 checks passed

Learns from every incident.

Notes live in a git repository you can read, so the next investigation starts ahead.

Settings · Memory
MEMORY.md
daily
preferences
repos
services
checkout.md
orders.md
search.md
services/checkout.md Edit
checkout

checkout-edge serves every checkout request on Cloudflare Workers.

How it works
•Reads orders through the Hyperdrive pool hd-prod.
•Peaks at 41 concurrent queries around 20:00 UTC.
Known failures
•Pool too small: requests queue and time out. Seen Nov 17 after deploy 9f3c2a1, fixed in #1318.
Updated by Polylane · 2 hours ago

Catches bad changes before they merge.

Polylane checks pull requests against what's running in production right now, whether a person or an agent wrote them.

How impact assessments work→
Pull request
Open coreplanelabs/orders-api #1290
perf(orders): add a trigram index for order search
polylane commented

Hold this merge. The orders table takes 1,900 writes a second at peak, and building this index without CONCURRENTLY would lock it for several minutes.

Tests 212 passed
Lint
Polylane production impact Hold this merge Required

Outcomes

Real numbers from Polylane customers.

Loved by developers

chronark @chronark just 2 hours after signing up for @polylane it prevented a potential problem in production. Jason @Jasonpraful When I installed polylane for the first time my expectations were pretty low. Couple weeks in I’m speechless - it just works everything is all of a sudden fully autonomous Kyle Galbraith @kylegalbraith fuuuuck @boristane and team cooked with polylane. already merged a pr. Dhravya Shah @DhravyaShah there was something wrong with our docs deployment, and Polylane automatically found and notified us about it... insane stuff @boristane Ben Anthony @benjamin_ACD a few days into using @polylane and I have merged like 40 something worthwhile PRs, seems my agents think the PRs are much higher utility than what detail was giving me Brenton West @west_brent34411 As I was testing Prod, started to notice things were not flowing like they did in Dev… I gave it a go. Issue Instantly ID. lucas @0xperp We started paying for polylane mostly bc of the handoff into our cursor account Ben @nurodev So far @polylane is very cool! Even more that it's finding issues, creating fixes and then working with @pullfrogai in PR's to improve it even more. Tito @mynameistito Wait... so you're telling me I don't even need to make PR's to fix up stuff and @polylane will just do it for me???? luna @ImLunaHey if you haven’t tried out @polylane at work i’d highly suggest it. tushar @gpu004 i really love polylane, such an awesome solution @boristane @polylane Sadok @yesadok polylane is so clean Xiaofeng Xie @XiaofengXie16 polylane is truly amazing Utpal Nadiger @utpalnadiger bullish on polylane
Boris Tane
Boris Tane Founder

I've spent my entire career on observability.

I founded Baselime, which Cloudflare acquired in 2024, then led Workers Observability at Cloudflare.

The tools kept getting better, and a person still had to read them and carry what they found back into the code. I started Polylane because nobody should be on-call.

Read why I'm betting my company on proactive agents→

Nobody should be on-call.