Gondolin vs Matchlock
Choose Gondolin for running agent-generated code in disposable local micro-VMs where egress, credentials and file access must stay under programmable host-side policy; choose Matchlock for running coding agents on a Linux server or Apple Silicon Mac in disposable microVMs with locked-down egress and API keys kept on the host. Gondolin uses libkrun / QEMU micro-VM isolation and is open source; Matchlock uses Firecracker / Apple VZ microVM isolation and is open source.
Key differences between Gondolin and Matchlock
- Type: Gondolin is a tool that runs on your own machine; Matchlock is an open-source project you run yourself.
- Isolation: Gondolin uses libkrun / QEMU micro-VM; Matchlock uses Firecracker / Apple VZ microVM.
- Licence: Gondolin is Apache-2.0; Matchlock is MIT.
- Self-hosting: Matchlock documents a self-hosted option; Gondolin does not.
- SDKs and interfaces: Only Matchlock lists Go and Python.
What they have in common
- Both use microVM isolation.
- Both are open source.
- Both list TypeScript and CLI.
Gondolin vs Matchlock at a glance
| Attribute | Gondolin | Matchlock |
|---|---|---|
| What it is | Apache-2.0 local Linux micro-VMs for agent-generated code, driven from a TypeScript SDK or CLI, with network egress, secrets and file mounts controlled by host-side JavaScript policy. | Experimental MIT CLI and Go, Python and TypeScript SDKs that run AI agents from OCI images in ephemeral microVMs, with a host-side proxy that enforces an egress allow-list and injects secrets the VM never sees. |
| Type | Local / OS-level | Open-source project |
| Isolation | Local Linux micro-VM per sandbox: libkrun where available (macOS on Apple silicon, Linux x86_64 with read/write /dev/kvm) and QEMU otherwise (Intel Macs, Linux arm64, Linux x86_64 without KVM via TCG); networking and the virtual filesystem run in a host daemon (gondolind) outside the guest | MicroVM per sandbox: Firecracker on Linux with KVM and Apple Virtualization.framework on macOS (Apple Silicon); no host directories are shared with the guest, and a host-side policy and HTTP/TLS proxy sits between the guest and the network |
| Open source | Yes (Apache-2.0) | Yes (MIT) |
| Self-host / BYOC | No / No | Yes / Not documented |
| SDKs | JavaScript/TypeScript, CLI | Go, Python, JavaScript/TypeScript, CLI |
| Cold start (vendor claim) | Docs: VMs "boot in under a second" on a Mac or Linux machine (vendor claim); no measurement methodology given | README: a full Linux environment that "boots in under a second" (vendor claim); pulled images are cached for faster startup |
| Max session | Not documented | Not documented |
| Persistence | Each VM starts from a clean copy of its image and the root disk is discarded on close by default (root disk modes readonly, memory or cow); disk-only checkpoints capture the writable root disk as a .qcow2 file that many VMs can resume from in parallel; tmpfs paths and VFS mounts are not in checkpoints, and durable data is meant to go to a host-backed mount such as /workspace | Guest-local changes are disposable unless a persistent block volume is attached (matchlock volume with run --disk @name:/guest/path). Sandboxes started with --rm=false or -d keep running for later exec. There are no host-directory mounts; files move in and out with matchlock cp, --copy-in/--copy-out or the SDK file APIs |
| Snapshots / pause | Yes / No | Not documented / Not documented |
| GPU | Not documented | Not documented |
| Networking | No NAT: the host daemon terminates guest traffic. HTTP/1.x and HTTPS are parsed and performed by the host (TLS intercepted with a local CA the guest trusts), with any public host allowed by default and loopback, private, link-local and metadata ranges blocked; other TCP and non-DNS UDP are dropped. allowedHosts / --allow-host restrict egress, and JavaScript hooks or daemon-side policies can filter and rewrite requests; secrets reach the guest only as placeholders that the host replaces for allowed hosts. HTTP/2, HTTP/3 and QUIC are not supported | Passing --allow-host or --secret seals the network so only allowed hosts are reachable, and the allow list can be edited at runtime; real secrets stay on the host and are injected into allowed outbound requests while the guest sees a placeholder. SDK interception rules can mutate requests and responses. The Go SDK blocks private IP ranges by default when a network config is sent; --no-network gives a guest with no NIC; ports can be published or forwarded. Linux uses a transparent proxy (nftables DNAT on ports 80/443); macOS uses Virtualization.framework NAT, or gVisor's userspace TCP/IP stack when intercepting |
| Pricing | Free (Apache-2.0) | Free and open source under MIT; no hosted offering is documented |
| Free tier | Free | Free (MIT) |
| Compliance | Not documented | Not documented |
| Regions | Not documented | Not documented |
| Company headquarters | Not documented | Not documented |
| Good fit for | Running agent-generated code in disposable local micro-VMs where egress, credentials and file access must stay under programmable host-side policy | Running coding agents on a Linux server or Apple Silicon Mac in disposable microVMs with locked-down egress and API keys kept on the host |
Verified November 18, 2026 from each provider’s documentation. Cold-start figures are vendor claims, not our measurements.
Which should you choose?
Reasons to choose Gondolin
- You want a microVM boundary around agent code on your own Mac or Linux machine without a cloud service
- You need egress allowlists and request hooks written in JavaScript rather than fixed profiles
- You want credentials usable by the guest without ever placing them inside it
- You want to prepare a base environment once and start many disposable VMs from a checkpoint
Reasons to choose Matchlock
- You want an agent to have a full Linux VM while its API keys never enter the guest.
- You need a default-deny egress allow-list you can edit while the sandbox runs.
- You run agents on both Linux servers and Apple Silicon Macs and want one CLI and SDK for both.
- You want no host-directory sharing, with explicit copy-in and copy-out.
Trade-offs to check first
- Gondolin: Docs describe it as an early project
- Gondolin: No full VM save/restore; checkpoints capture the root disk only
- Matchlock: Labelled experimental and subject to breaking changes.
- Matchlock: No host-directory mounts; files must be copied in and out or kept on block volumes.
Both are worth piloting with your own image and region. See the full Gondolin profile and Matchlock profile, or the complete comparison.
Frequently asked questions
Is Gondolin or Matchlock better for AI agents?
What is the difference between Gondolin and Matchlock?
Which is open source, Gondolin or Matchlock?
How do Gondolin and Matchlock isolate code?
How does Gondolin pricing compare with Matchlock?
Can you self-host Gondolin or Matchlock?
Do Gondolin and Matchlock support snapshots or pause and resume?
Sources
- earendil-works.github.io/gondolin/
- earendil-works.github.io/gondolin/quickstart/
- earendil-works.github.io/gondolin/workloads/
- earendil-works.github.io/gondolin/snapshots/
- github.com/jingkaihe/matchlock
- github.com/jingkaihe/matchlock/blob/main/AGENTS.md
- github.com/jingkaihe/matchlock/tree/main/docs
- github.com/jingkaihe/matchlock/blob/main/docs/lifecycle.md
Check before you rely on it. Products change quickly, and this page reflects public documentation as of November 18, 2026. Confirm anything that matters in the provider’s own documentation.
Cite this page
Agent Sandboxes. (2026). Gondolin vs Matchlock. agentsandboxes.org. Last reviewed November 18, 2026. https://agentsandboxes.org/compare/gondolin-vs-matchlock/