Koyeb vs AWS AgentCore
Koyeb runs each sandbox in its own microVM from any Docker image, which can sleep when idle and wake on the next request, on CPU or GPU. AgentCore is the part of AWS that hosts agents in 2 vCPU microVM sessions of up to 8 hours, or on EC2 Instances for up to 14 days, with managed Code Interpreter and Browser tools inside your AWS account.
The companies and products
Koyeb
Koyeb is a serverless cloud for apps, inference and sandboxes. In February 2026 it agreed to join Mistral AI.
A Koyeb sandbox is a Koyeb service: a microVM that starts from any Docker image. You drive it from Python or JavaScript SDKs or the Koyeb CLI: run commands, read files, expose a port and take snapshots. The default image comes with four coding agents. A sandbox can scale to zero when idle and wake on the next request, on CPU or GPU instance types. It suits teams that run AI agents or untrusted code next to apps they already host on Koyeb.
AWS AgentCore
Amazon Bedrock AgentCore is a set of AWS services to deploy and run AI agents: a Runtime that hosts your agent code, plus built-in Code Interpreter and Browser tools.
AgentCore is not a sandbox product first. It is a hyperscaler's agent platform: you push a container image or code, and each Runtime session gets its own Firecracker microVM. You reach a session through AgentCore APIs, an interactive shell API and authenticated HTTPS session endpoints. Around it sit managed tools (Code Interpreter, Browser) and services for memory, identity, gateways and evaluations. An Instances compute type runs longer sessions on EC2 in your own account. It is built for teams who already build on AWS and want agents to stay inside their account, IAM and VPC.
Which use cases each one fits
For each use case, Battleships prices every sandbox provider on the same workload and sorts them from the cheapest monthly bill up. “6th cheapest of 336” means 5 of the other 335 providers run that workload, with every feature it needs, for less. “With limits” means the provider runs it but misses something, so it is listed after every full match. Open a use case for its whole ranking and what each provider misses.
| Use case and its workload | Koyeb: price rank | AWS AgentCore: price rank | boat: price rank |
|---|---|---|---|
| Software factory 100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month Both are compromises for a software factory: Koyeb documents neither Docker inside a sandbox nor nested virtualization (65th), and AgentCore keeps files only under /mnt (95th). boat ranks 8th with a full VM that saves its whole disk. | 92nd of 365, with limits $366 a month | 132nd of 365, with limits $888 a month | 5th cheapest of 365 $634 a month |
| Always-on AI employees 50 machines 24/7 · 2 vCPU / 4 GB An assistant that runs all month outlasts AgentCore's 8-hour microVM sessions (74th, compromise). On Koyeb you take an SDK snapshot before Deep Sleep discards the VM, or keep it awake (123rd, compromise). boat ranks 5th. | 194th of 361, with limits $923 a month | 109th of 361, with limits $1,300 a month | 6th cheapest of 361 $657 a month |
| Code interpreter bursts 100 at once · 1 vCPU / 2 GB · 200,000 sessions of 1 min a month Both fit short code runs: Koyeb ranks 5th at $66 a month with small instances that scale to zero, and AgentCore 9th at $114 with a Code Interpreter billed only for CPU actually used. boat's start limits make big bursts a compromise (130th). | 6th cheapest of 336 $66 a month | 11th cheapest of 336 $114 a month | 189th of 336, with limits $2,000 a month |
| Sandboxes for your users 300 at once · 2 vCPU / 4 GB · 30,000 sessions of 23 min a month Koyeb ranks 6th: each user's sandbox can expose an HTTPS port and wake on request. AgentCore sessions expose 1 to 5 ports only through an authenticated endpoint (IAM or OAuth on every request), so it is a compromise (127th). boat ranks 2nd, with up to 50 HTTPS URLs each. | 8th cheapest of 349 $495 a month | 205th of 349, with limits $506 a month | 3rd cheapest of 349 $209 a month |
| RL rollouts 1,000 at once · 2 vCPU / 4 GB · 1,000,000 sessions of 3 min a month | 233rd of 349, with limits $1,631 a month | 129th of 349, with limits $8,075 a month | 39th of 349, with limits $2,000 a month |
| Agent evals 250 at once · 4 vCPU / 8 GB · 15,000 sessions of 20 min a month | 27th cheapest of 359 $429 a month | 46th cheapest of 359 $650 a month | 10th cheapest of 359 $200 a month |
| CI runners 40 at once · 8 vCPU / 16 GB · 6,000 sessions of 12 min a month AgentCore fits CI runners on its Instances type, at $399 a month (34th). Koyeb's fitting price here is a negotiated reservation (104th, compromise). Neither documents /dev/kvm for Android or VM builds. boat ranks 2nd at $86. | 143rd of 374, with limits $88 a month | 51st cheapest of 374 $399 a month | 2nd cheapest of 374 $86 a month |
| 24/7 dev boxes 1 machine 24/7 · 4 vCPU / 8 GB | 232nd of 371, with limits $37 a month | 242nd of 371, with limits $52 a month | 19th cheapest of 371 $26 a month |
| Computer-use agents 50 at once · 4 vCPU / 8 GB · 20,000 sessions of 15 min a month Computer-use agents need a desktop. On Koyeb you add a desktop and VNC to your own image (80th), and AgentCore Browser drives a browser window only (95th); both are compromises. boat ranks 8th with a 60 fps Linux desktop. | 128th of 367, with limits $123 a month | 146th of 367, with limits $202 a month | 7th cheapest of 367 $200 a month |
| Browser-use agents 50 at once · 2 vCPU / 4 GB · 30,000 sessions of 10 min a month AgentCore ranks 4th at $144 a month with a managed Chromium that has live view and replay. On Koyeb you add a browser to your own image (142nd, compromise). boat ranks 6th at $200 with Chrome on a full desktop. | 237th of 356, with limits $123 a month | 4th cheapest of 356 $144 a month | 7th cheapest of 356 $200 a month |
| Scraping & bots (startup scale) 50 at once · 1 vCPU / 2 GB · 100,000 sessions of 3 min a month | 242nd of 343, with limits $89 a month | 32nd cheapest of 343 $287 a month | 6th cheapest of 343 $90 a month |
| Mass scraping & bots (undetectable) 1,000 at once · 1 vCPU / 2 GB · 3,000,000 sessions of 2 min a month | 67th of 327, with limits $1,683 a month | 164th of 327, with limits $5,458 a month | 73rd of 327, with limits $1,800 a month |
| Parallel red teaming 200 at once · 2 vCPU / 4 GB · 5,000 sessions of 30 min a month | 261st of 351, with limits $299 a month | 108th of 351, with limits $294 a month | 3rd cheapest of 351 $200 a month |
| GPU eval jobs 10 at once · 8 vCPU / 32 GB · 500 sessions of 30 min a month Koyeb ranks 3rd, with GPU instances from an RTX 4000 SFF Ada up to H100, H200 and B200. AgentCore offers GPUs on its Instances type in your account. boat has no GPU. | 3rd cheapest of 102 $628 a month | Not ranked: misses a requirement | Not ranked: misses a requirement |
| macOS agents 5 at once · 4 vCPU / 8 GB · 1,000 sessions of 20 min a month | Not ranked: misses a requirement | Not ranked: misses a requirement | Not ranked: misses a requirement |
Key features
| Feature | Koyeb | AWS AgentCore | boat |
|---|---|---|---|
| Isolation | Cloud Hypervisor microVM per instance, started from any Docker image | Firecracker microVM per Runtime session; Code Interpreter and Browser in a containerized environment | Full VM with its own kernel, on physical hosts dedicated to boat (AMD Ryzen 7 9700X and Ryzen 9 9950X); vCPUs are shared |
| Biggest machine | Instance types up to 40 vCPU / 128 GB, and larger GPU types | 2 vCPU / 8 GB per microVM session, not adjustable; on Instances, the EC2 type you pick sets the size | Up to 16 vCPU / 32 GB with a 251 GB disk, on the $100 plan and above |
| Session length | No forced cap documented; optional auto-delete timers up to 24 hours | Up to 8 hours per lifecycle on microVMs, 15 idle minutes by default; up to 14 days on Instances | Auto-stop defaults to 1 hour; set up to 30 days, or turn it off on a paid plan for no session cap |
| What survives a stop | Local disk is lost on redeploy and in Deep Sleep; SDK snapshots save the filesystem for new sandboxes; volumes (preview, 1 to 10 GB, not on eco instances) persist | Mounts under /mnt only: a 1 GB per-session storage mount (preview), your own EFS or S3 Files mounts in a VPC, or EBS volumes on Instances; the rest resets | Whole disk saved every minute and at each stop; kept at least 14 days after a stop; memory and running programs are not saved |
| Snapshots and fork | SDK snapshots of the filesystem, or with running processes, then new sandboxes from them | No snapshot or fork of a session; on Instances you can seed new EBS volumes from your own EBS snapshots | Automatic disk snapshot every minute; fork a running sandbox from its last 1-minute snapshot (no memory) |
| Sleep and wake | Scales to zero after 5 idle minutes or more and wakes on the next request; Light Sleep wakes in about 200 ms (preview); not on eco instances | Session stops after 15 idle minutes by default; the next call starts a fresh microVM | No wake on request; stop and start by API; no compute charge while stopped |
| Docker, systemd and /dev/kvm | Starts from any Docker image; Docker inside a sandbox and nested virtualization not documented | Root inside the microVM; Docker and /dev/kvm not documented | Root, Docker, systemd and /dev/kvm for Android emulators and VMs; you cannot boot your own Docker image |
| Inbound access | One exposed HTTPS port at a time, plus an optional TCP proxy to one port (preview); no dedicated IP | Authenticated HTTPS session endpoint for 1 to 5 ports (IAM or OAuth on every request, no public links); interactive shell API; no SSH or raw TCP | Up to 50 HTTPS URLs, token-gated by default; SSH; public IPv6; the IP changes at each resume |
| Network controls | Block all outbound traffic, or allow only listed IPs and CIDRs | VPC mode with security groups, PrivateLink and a NAT gateway for a fixed outbound IP | Open internet, no provider-side allowlist |
| Browser and desktop | No desktop or browser preinstalled; you add them to your own Docker image | Managed Chromium with live view, session recording and replay, profiles and proxies; no desktop beyond the browser window | 1920x1080 Linux desktop and Chrome at 60 fps, with computer-use tools |
| Coding agents | Default image preinstalls Mistral Vibe, Codex, Gemini and OpenCode, run through the command API; no prompt or steer API | AWS's own managed harness built on Strands Agents, or your own agent code; you set up other harnesses in your container image | Seven harnesses installed, including Claude Code and Codex, with a prompt, steer and events API |
| Keeping API keys out of the sandbox | Secrets are set as environment variables inside the service | AgentCore Identity stores OAuth tokens and API keys and hands them to your agent code | Optional AI credentials proxy ($1 a day on, plus $1 per 1,000 requests): real keys of supported providers stay out; other secrets still enter |
| GPUs | RTX 4000 SFF Ada up to H100, H200 and B200, and multi-GPU | NVIDIA G-series and Inferentia2 on the Instances type | None on any size |
| Regions | EU, US and Asia; you pick the region | 21 commercial AWS regions plus GovCloud (Instances in 9) | Germany, Finland and France; you cannot pick a region; no US or Asia region |
Every feature Battleships tracks (43)
| Feature | Koyeb | AWS AgentCore | boat |
|---|---|---|---|
| Full VM (own kernel) | Yes | Yes | Yes |
| Docker inside | Not published | Not published | Yes |
| Nested virtualization | No documented way | No documented way | Yes |
| Root | Not published | Yes | Yes |
| Systemd | Not published | No documented way | Yes |
| Desktop GUI | No documented way | No documented way | Yes |
| Browser | No documented way | Yes | Yes |
| Computer-use API | No documented way | Yes | Yes |
| SSH | No documented way | No documented way | Yes |
| Persistent disk | With your own setup | Yes | Yes |
| Pause/resume | Yes | No documented way | Yes |
| Automatic snapshots | Yes | No documented way | Yes |
| Fork/clone | Yes | No documented way | Yes |
| Live fork (no pause) | Not published | No documented way | Yes |
| Memory fork | Yes | No documented way | No documented way |
| Custom image (Docker or snapshot) | Yes | Yes | Yes |
| ≥24 h sessions | Yes | No documented way | Yes |
| Idle auto-stop | Yes | Yes | No documented way |
| Wake on request | Yes | Yes | No documented way |
| HTTPS preview URLs | Yes | Yes | Yes |
| Raw TCP inbound | Yes | No documented way | Yes |
| Public IPv4 | No documented way | No documented way | No documented way |
| Egress allowlist | Yes | Yes | No documented way |
| Secret proxy | No documented way | No documented way | Yes |
| Secret proxy for any API | No documented way | No documented way | Yes |
| Preinstalled agents | Yes | No documented way | Yes |
| Agent harness API | No documented way | No documented way | Yes |
| MCP server | Yes | Yes | No documented way |
| GPU | Yes | Yes | No documented way |
| Windows | No documented way | No documented way | No documented way |
| Arm64 | No documented way | Yes | No documented way |
| Self-hosting | Yes | No documented way | No documented way |
| BYOC | Yes | Yes | No documented way |
| Open source | No documented way | No documented way | No documented way |
| Scoped API keys | Not published | Not published | Yes |
| Spend limits | No documented way | With your own setup | Yes |
| Usage per sandbox (API) | No documented way | No documented way | Yes |
| Audit logs | Yes | Yes | No documented way |
| SSO | Yes | Yes | No documented way |
| SOC 2 | Yes | Yes | No documented way |
| HIPAA | No documented way | Yes | No documented way |
| EU data residency | Yes | Yes | No documented way |
| Zero data retention | Not published | Not published | Yes |
Performance
Koyeb and AgentCore are not in the public benchmarks we use, so we cannot say which is faster. In StarSling's run of 3 October 2026 (run 37073680030), boat did 28.06 runs/s on the V8 Web Tooling Benchmark at 4 vCPU / 8 GB, second of 15. Every boat sandbox in that run ran on an AMD Ryzen 9 9950X host. boat also places some sandboxes on AMD Ryzen 7 9700X hosts, which this run did not measure. Koyeb publishes a CPU share of at least 0.25 dedicated vCPU per GB of memory, and an AgentCore microVM session stops at 2 vCPU, so a like-for-like test there would need the Instances type.
V8 Web Tooling Benchmark at a 4 vCPU / 8 GB target, runs per second, higher is better. Source: StarSling's run of 3 October 2026 (run 37073680030), median of 5 sandboxes. Providers ran on different hardware in different regions. StarSling gives a shared rank when the run cannot separate two providers. Every boat sandbox in this run ran on an AMD Ryzen 9 9950X host. boat also places some sandboxes on AMD Ryzen 7 9700X hosts, which this run did not measure. Koyeb and AWS AgentCore are not in the public benchmarks we use, so we cannot say which is faster.
Pricing in five scenarios
Each bill is priced from the providers' public price lists by the open Battleships engine, with the cheapest plan that runs the scenario. Start limits, per-start fees and plan minimums are in the price, boat's included.
| Scenario | Koyeb, a month | AWS AgentCore, a month | boat, a month |
|---|---|---|---|
| One coding agent, workdays 4 vCPU / 8 GB, 50 GB disk, 8 hours a day, 22 days a month, files kept between days | $29 · Pro | $26 · Pay-as-you-go (Instances) · with limits | $20 · $20/mo |
| One agent that never stops 4 vCPU / 8 GB, 50 GB disk, running 24/7 all month | $57 · Pro | $107 · Pay-as-you-go (Instances) · with limits | $26 · $20/mo |
| 50 agents in parallel 50 machines of 4 vCPU / 8 GB at once, 8-hour sessions, 1,100 sessions a month, Docker inside | $475 · Pro · with limits | $1,286 · Pay-as-you-go (Instances) · with limits | $317 · $20/mo × 3 seats |
| Sandboxes for your users 2 vCPU / 4 GB, 30,000 sessions of 10 minutes a month, up to 300 at once | $329 · Scale | $423 · Pay-as-you-go · with limits | $200 · $100/mo × 2 seats |
| 1,000 short jobs at once 2 vCPU / 4 GB, 1,000 at once, 3 minutes each, 200,000 jobs a month | $485 · Scale · with limits | $1,615 · Pay-as-you-go (Instances) | $8,800 · $2000/mo · with limits |
Koyeb: Koyeb bills each instance per second at a fixed price per instance type: eco-xlarge (4 vCPU / 8 GB, 20 GB disk) costs $0.0576 an hour. The idle period before sleep bills at the full rate, and Deep Sleep has no charge. Pro costs $29 a month with $10 of usage included, and Scale $299 with $100 included; the rest of the fee is not credit.
AWS AgentCore: No monthly fee. Runtime v1 microVMs, Code Interpreter and Browser bill per second: $0.0895 per vCPU-hour for CPU actually used and $0.00945 per GB-hour for peak memory. The 4 vCPU shape in these scenarios is the Instances type: the EC2 on-demand price plus a 12% management fee. EBS storage, billed also while a session is stopped, and data transfer bill on top; the scenario prices leave out EBS.
AWS AgentCore: Plans cost $20, $100, $500 or $2,000 a month, and the whole plan price comes back as VM time for that month. A 4 vCPU / 8 GB sandbox costs $0.036 an hour, billed per second while it runs, busy or idle. The $20 plan includes 12 starts a minute, 60 an hour and 200 a day; extra starts cost $0.03 to $0.07. A stopped sandbox has no compute charge and keeps its disk for at least 14 days; saved snapshots above the included count and the AI credentials proxy still charge. A card is required, also for the 7-day trial.
Limits in those scenarios
AgentCore's microVM sessions stop at 2 vCPU / 8 GB, so the 4 vCPU scenarios use its Instances type, where files survive a stop only under /mnt and a session lasts up to 14 days. So AgentCore does not fit the agent scenarios as a work machine. Koyeb's priced eco-xlarge shape has a 20 GB disk and takes no volume, and Koyeb does not document Docker inside a sandbox, so the 50-agent fleet is not confirmed. For 1,000 short jobs at once, AgentCore fits at $1,615 a month. boat does not fit that burst: its $2,000 plan includes 90 starts a minute, charges for extra starts and refuses starts above 270 a minute.
| Scenario | Koyeb | AWS AgentCore | boat |
|---|---|---|---|
| One coding agent, workdays |
|
| None for this scenario |
| One agent that never stops |
|
| None for this scenario |
| 50 agents in parallel |
|
| None for this scenario |
| Sandboxes for your users | None for this scenario |
| None for this scenario |
| 1,000 short jobs at once |
|
|
|
Which one should you use?
Use Koyeb when
your sandboxes sleep most of the time and must wake on a request, start from your own Docker image, need a GPU or a big instance, or should expose a port to your users.
- Bigger machines and GPUs without your own cloud account. Koyeb sells instances up to 40 vCPU / 128 GB and GPUs up to H100, H200 and B200, billed per second.
- Sleep when idle, wake on the next request. For sandboxes idle between requests: Koyeb scales to zero after 5 minutes and wakes in about 200 ms with Light Sleep (preview).
- A port your users can open. A Koyeb sandbox can expose an HTTPS port and a TCP proxy, so a user can open the app it runs.
Use AWS AgentCore when
your agents must stay inside your AWS account and VPC, need one of its 21 commercial regions, or are short code and browser tasks that fit its managed tools.
- It lives inside your AWS account. Teams built on AWS keep agents behind their own IAM, VPC, security groups and PrivateLink, in 21 commercial regions plus GovCloud.
- Managed Code Interpreter and Browser. For short code runs and browser tasks, both tools bill only for CPU actually used and peak memory, and Browser adds live view and replay.
- Two-week sessions on Instances. Agents that need a GPU or up to 14 days can run on EC2 in your account, with Savings Plans applying.
Where boat fits
Koyeb and AgentCore both run short, isolated work well, and both need a step from you to keep files: a snapshot or volume on Koyeb, a /mnt mount on AgentCore. If your agents work for hours on one machine, installing packages, running Docker and serving what they build, boat gives each one a full VM that saves its disk with no code, with agents, a desktop and ports installed.
A disk saved with no code
On Koyeb you take SDK snapshots or attach a preview volume of up to 10 GB. On AgentCore only paths under /mnt survive. boat saves the whole disk every minute and at each stop, and can fork a running sandbox. A fork can miss the last minute of writes.
A full machine with /dev/kvm
AgentCore microVM sessions stop at 2 vCPU and 8 hours unless you run EC2 Instances in your own account. Neither documents /dev/kvm. boat gives root, Docker, systemd and /dev/kvm, up to 16 vCPU.
boat's own limits
boat has no GPU, no wake on request, no outbound allowlist and no US or Asia region. A sandbox stops after 1 hour by default unless you change auto-stop. Starts are capped per plan, so 1,000 short jobs at once fit AgentCore but not any boat plan.
Use boat when your agents do real work on one machine for hours: installs, builds, Docker, emulators and a desktop browser, with a disk saved every minute and HTTPS URLs you can open. Compare it one to one: boat vs Koyeb and boat vs AWS AgentCore.
boat's own limits
- Plans start at $20 a month. The plan price comes back as VM time for that month. Unused time does not carry over. Source
- Starts (create, fork, resume): the $20 plan includes 12 a minute, 60 an hour and 200 a day. The $2,000 plan includes 90 a minute, 600 an hour and 2,400 a day. Extra starts cost $0.03 to $0.07. A start fails with 429 above 3 times the included starts per minute, or 6 times per hour or per day. Source
- A stopped sandbox has no compute charge. boat keeps its disk for at least 14 days. A saved snapshot keeps it longer; extra saved snapshots and the AI credentials proxy are paid. Source
- Auto-stop defaults to 1 hour. You can turn it off on a paid plan. The trial allows 2 hours at most. Source
- boat saves the disk every minute and at each stop. Memory, running programs and /tmp are not saved: a resume is like a reboot. A machine failure loses writes after the last good snapshot. Source
- Machines run in Germany, Finland and France only. Snapshot copies can be stored outside the EU. Source
- Public IPv6 and outbound IPv4. No dedicated IPv4. The IP changes at each resume. Source
- Hosts are AMD Ryzen 7 9700X and Ryzen 9 9950X. vCPUs are shared. No GPU. You cannot boot your own Docker image. Source
Questions people ask
Is Koyeb cheaper than AWS AgentCore?
For long-lived agents, yes. One agent that never stops costs $57 a month on Koyeb, on a 20 GB disk, and $107 on AgentCore, which does not fit as a work machine. For short code runs, Koyeb comes to $66 and AgentCore to $114. For 1,000 short jobs at once, AgentCore fits at $1,615 a month. boat comes to $26 for the always-on agent.
Which is faster, Koyeb or AgentCore?
Neither is in the public benchmarks we use, so we cannot say. A single AgentCore microVM session stops at 2 vCPU, while Koyeb sells larger instances. In StarSling's run of 3 October 2026, boat did 28.06 runs/s at 4 vCPU / 8 GB, on a Ryzen 9 9950X host.
How long can a session run on each?
An AgentCore microVM session lasts up to 8 hours and stops after 15 idle minutes by default; the Instances type allows up to 14 days. Koyeb documents no forced cap, only optional auto-delete timers. A boat sandbox stops after 1 hour by default; on a paid plan you can turn auto-stop off for no session cap.
Do files survive a stop on Koyeb or AgentCore?
Yes, with a step from you. On Koyeb you take an SDK snapshot of the filesystem or attach a volume of up to 10 GB. On AgentCore you keep files under /mnt: a 1 GB storage mount in preview, your own EFS or S3 Files, or EBS on Instances. boat saves the whole disk every minute and at each stop; memory, running programs and /tmp are not saved.
Can Koyeb or AgentCore run Docker and Android emulators?
Koyeb starts each sandbox from a Docker image, and AgentCore runs your container image as root, but neither documents Docker inside a session or /dev/kvm, which Android emulators and VMs need. boat runs Docker with root and systemd and exposes /dev/kvm.
Do Koyeb and AgentCore have GPUs?
Koyeb runs GPU instances from an RTX 4000 SFF Ada up to H100, H200 and B200, billed per second. AgentCore offers NVIDIA G-series and Inferentia2 on the Instances type in your AWS account. boat has no GPU on any size.
Try boat
Install the CLI, then run boat new.
7-day free trial