The AWS Lambda console is good at one thing: looking after one function, in one Region, in one account. That is how most teams start. It is not how most teams end up. A few years in, there are hundreds of functions spread across development, staging and production accounts, in two or three Regions, owned by different teams, and the questions people ask have changed:
- Which of our functions are failing right now, everywhere?
- Which ones has nobody called in months?
- Who changed this function yesterday, and was it a deployment or a console edit?
- Can the payments team test their own functions without getting access to everything else?
The console can answer some of these, one function at a time. FunctionPilot is our answer to all of them at once. It is in early access now.
One table for every function
FunctionPilot opens on a single table of every Lambda function across the AWS accounts and Regions you connect. Each row carries the numbers the console's function list leaves out: invocations, errors and error rate, throttles, average, p95 and maximum duration, peak concurrency, a sparkline and an estimated cost, for any timeframe from 15 minutes to 30 days.
You can sort by any column, so "the functions with the most errors this week" is one click. Filters cover account, Region, runtime, tags, deprecated runtimes and functions with no invocations. A Last invoked column, worked out from 15 months of daily metrics, makes unused functions obvious, and an "unused for N days" filter turns that into a clean-up list. Views with their filters, columns and sort can be saved for yourself or shared with the team.
A function page that answers the next question
Open a function and its page brings together what you would otherwise collect from four consoles:
- Overview: metric charts on one timeframe, with every deployment from CloudTrail drawn on them as a marker, so a jump in errors lines up with the change that caused it. Console edits are flagged.
- Code: the deployment package of any version or alias, file by file, a side by side diff between two versions, and an editor for small fixes (with a clear warning that it bypasses your pipeline).
- Configuration: everything in readable cards, with environment variables that look like secrets flagged and a suggestion to move them to Secrets Manager.
- Logs: the function's own log group, opened for you, with filters, levels, request ids and lines grouped by invocation.
Tests your team can share
The test workbench keeps saved tests in FunctionPilot rather than in the console's ten private events per function. A test can be private or shared with everyone who may invoke the function, and every run keeps its response, log tail, billed duration, cold start and memory used. Assertions on the status, JSONPath values or the duration turn a saved test into a check, and Run all tests runs every test of a function in one go.
Access the console can't give
Admins grant users and groups view, logs, code, invoke, edit or manage on functions by name pattern, per account and Region: the payments team can test payments-* in staging and only read production. AWS enforces those grants itself, through a session policy on every request, and every call reaches CloudTrail with the person's email as its source identity. Teams on IAM Identity Center can also work with their own permission sets instead.
In your account, with nothing leaving it
FunctionPilot is one CloudFormation stack in your own account: a serverless API, a web app on Amplify Hosting and Cognito sign-in through IAM Identity Center, SAML or OIDC. Other accounts connect through a small role. Nothing is sent to us, and the parts of CloudWatch that cost money per call are cached or worked out on a schedule, so opening the table a hundred times doesn't mean paying for it a hundred times.
FunctionPilot will be sold on AWS Marketplace. If you would like early access, get in touch (opens in a new tab) and tell us how many functions and accounts you run.