AWS, checked without opening the console.
Ask what is running, what the logs say and what the bill is doing, and Cedric proposes the restart, the resize or the snapshot before touching anything.
100K credits ($50) on sign up. No card needed.
- For a small team the console is where you go when something is wrong, and the question is usually what is running and why it is slow or expensive.
- Ask in plain words: the servers and containers in each region with their health, the alarms that fired, the logs around an error.
- Cedric can start, stop, restart or resize an instance, trigger a database snapshot, or invoke a function with test input, each proposed first.
- He reads cost and usage summaries and budgets, and names which services are driving spending this month.
- Cedric can only do what the connected credentials allow, and terminating servers, deleting storage or changing permissions always waits for your yes.
What can an AWS AI agent do?
Most of what a small team needs from AWS is a status check. An AWS AI agent gives it from the conversation. Cedric lists and describes the running servers and containers, with their configuration, region and health, so "what is running in eu-west-1 right now" has a short answer. He pulls performance metrics, alarms and application logs to investigate an error or a slowdown, and he inspects managed databases and caches to show their status and settings. When a function misbehaves, Cedric can view its configuration and invoke it with test input to see what comes back.
Changing things is proposed first. Cedric can start, stop, restart, resize or terminate compute instances and scale groups, trigger a snapshot or backup on a database, and browse storage buckets and files, upload or download an object, or check and change its access settings. He also handles access itself: listing users, roles and permission policies, rotating keys, and reviewing who can do what. A change to permissions or a terminate is the kind of write Cedric always shows in full before asking for your yes.
The bill is part of the same conversation. Cedric reviews cost and usage summaries and budgets and names which services are driving spending, which is the question behind most weekly reports an operator writes by hand. Everything runs as the connected credentials, so Cedric sees and does exactly what they allow, and calls are rate limited per service and region.
The slow morning
One alarm, one log line, one restart proposed.
Sofia Andersson9:14 AM
@Cedric the API has been slow since about 8am. What's running for the api service in eu-west-1, did any alarm fire, and what do the logs say?
CedricAPP9:15 AM
Two instances behind the api group in eu-west-1: api-1 (healthy, CPU 31%) and api-2 (health check failing since 08:04, CPU 97%). One alarm fired at 08:06: api-2 CPU above 90% for 5 minutes. Logs on api-2 from 08:03 repeat "connection pool exhausted" every few seconds; api-1 shows nothing unusual. Proposed: restart api-2 and keep api-1 as is. The group would run on one instance for about two minutes.
Described
Running servers and containers by region with configuration and health, plus databases, caches and functions with their settings.
Investigated
Metrics, alarms and application logs around an error or a slowdown, and a function invoked with test input to reproduce it.
Adjusted
An instance started, stopped, restarted or resized, a snapshot triggered, a key rotated or a bucket's access changed, on your yes.
Connect AWS with your own key
No setup project, no workflow builder, no field mapping. The flow is the same for every tool Cedric connects to.
- 01
Install Cedric where your team talks
One click. He joins your Slack like any other member, and everyone shares the same Cedric.
- 02
Connect AWS from the dashboard
Open the Integrations page, pick AWS and paste an access key pair from AWS IAM. You can revoke the connection in one click at any time.
Pick Read-only when you connect if you only want him reading. You can switch it later.
- 03
Ask in plain words
Back in the channel, try it: "what is running right now and did any alarm fire today?" You describe the outcome; Cedric picks the AWS actions the ask needs. There is nothing to configure.
Connect for the whole workspace so the team shares it, or keep the connection personal. Two accounts are fine too; each gets a short name.
- 04
Let it become routine
Schedule the recurring version of the ask, and let Cedric fold AWS into bigger jobs that cross your other tools. Anything that would leave your workspace still comes back as a preview to approve first.
Better in company
A slowdown seen in AWS logs usually has a deploy behind it. Cedric can read the alarm and the log lines, find the matching pull request in GitHub, and file the follow-up in Linear with both linked, each write on your yes.
The monthly cost summary belongs next to revenue. Ask Cedric for spending by service from AWS and the month's collected payments from Stripe, and he can put both in a Google Sheets tab without anyone exporting.
Worth knowing
AWS comes from Cedric's catalog of 3,200+ apps. He works from what the AWS connection can read and do, the same list its card shows in your dashboard.
Only what the credentials allow. Cedric acts as the connected credentials, so a resource they cannot see stays invisible and a change they are not permitted to make stays undone.
Terminating, deleting and permission changes are confirmed every time. Cedric shows the exact instance, bucket or policy before he acts, and calls are rate limited per service and region.
The AWS connection sets the scope. Cedric sees what it grants. Access is logged, and you can revoke it from the dashboard at any time.
Nothing sends itself. An email, a post, a change to a record: anything with consequences comes back as a preview for a person to approve.
FAQ
Yes. AWS is one of the 3,200+ apps in Cedric's catalog. Connect it once from the dashboard, with an access key pair from AWS IAM, and ask in plain words. Cedric picks the actions an ask needs from what the connection can read and do, with no fields to map.
No. Stopping, resizing or terminating an instance is a proposal that names the instance, the region and the effect, and nothing happens until you say yes. Reading status, metrics and logs never needs a confirmation.
Yes. He reads cost and usage summaries and budgets and names which services are driving spending, so he can say that storage grew or that a second instance ran all month. He reports the numbers AWS provides; he does not estimate beyond them.
Connect it for the whole workspace and everyone shares it, with one person authorising once. A connection you would rather not share, a personal inbox for instance, can stay personal to you.
Yes, and it starts that way: a key connection is read-only until you turn writes on. Even with writes on, anything that changes or sends waits for an approval.
Every workspace starts with 100,000 free credits (worth $50) on sign-up, no card. Plans start at $19 a month for the whole team, no seats. Each task uses credits in proportion to the work. See pricing.
Works well with AWS
CloudflareDNS, cache and firewall read and adjusted from the thread, every live change on your yes.
GitLabIssues, merge requests and pipelines read and moved from the thread, every merge on your yes.
SentryErrors found, stack traces read, issues triaged and assigned from the thread, on your yes.
DatadogMetrics, logs and monitors read in the thread; alerts muted and monitors changed on your yes.
PagerDutyIncidents, on-call and overrides read and handled from the thread, every page on your yes.
AsanaAssignees asked, tasks updated from their answers, and the project current before every meeting.
Connect AWS in a minute.
Ask what is running and what the bill is doing, then see what he proposes for the next alarm.
100K credits ($50) on sign up. No card needed.