Skip to main content
Check status.anonalabs.com first: if something is down there, it is not your code.
  • Check that ANONA_API_KEY (or E2B_API_KEY, for the stock e2b SDK) holds an Anona Compute key from the dashboard. E2B keys and Anona Memory keys do not work here.
  • The key may be revoked or expired; see API keys in the dashboard.
  • Keys are ac_ followed by lowercase hex; older keys start with e2b_ and still work. Check for a stray space or newline when you copy it.
Set E2B_DOMAIN=compute.anonalabs.com. Recent SDKs ignore E2B_API_URL. Or use anona_compute.Sandbox / @anona-labs/compute-sdk, which default to the Anona domain.
  • The sandbox may have reached its timeout and been deleted (the default is 300 s from the SDK). Use a longer timeout, set_timeout, or lifecycle={"on_timeout": "pause"}.
  • The key may belong to another organization, or be limited to a list of sandbox IDs. Both look exactly like “not found”.
Your key lacks sandboxes:use (needed for commands, live view and SSH), or your dashboard role is viewer. Create a key with Create and use, or ask an admin for the member role.
anona-compute usage, the MCP get_usage tool and GET /v1/usage need a key with usage:read (the Read only and Full access presets have it); billing needs billing:read (Full access). Create a new key; existing keys keep their scopes.
Team egress settings and the credential vault are changed by dashboard owners and admins, or by an unrestricted API key. Any role, and any key without a sandbox list, can read them.
The rule needs a new binding, and you may not make one. A binding belongs to one sandbox, so this happens even when another sandbox has the same rule. Create the sandbox as an organization owner or admin, or with an API key that has the vault:bind scope (a custom key; no preset has it). Or ask an owner to put the rule in the team egress settings, which every sandbox can use. See Bindings and who can make them.
Store the secret under exactly the name the rule uses (${vault.<name>}), then create the sandbox again. Either the dashboard’s Credential vault page, or:
store_secret.sh
Every secret records the hosts it may be sent to, and a rule aiming it anywhere else is refused. Without that list, anyone who can write an egress rule could point your credential at a server they control. Check the list with anona-compute vault list and replace it with anona-compute vault allowed-hosts <name> --host api.github.com. A *. entry does not cover the domain itself: *.github.com does not cover github.com.A secret stored before allowed hosts existed has none recorded and cannot be used until you set one.
The egress proxy could not resolve the secret, so it refused the request instead of sending it without the credential. That is deliberate: a request that silently lost its Authorization header would reach the upstream unauthenticated.The handshake to the host still succeeds, so this looks like the upstream failing rather than the policy. Causes, commonest first:
  • The secret was deleted, or the binding was revoked (anona-compute vault list shows neither).
  • The allowed-hosts list was narrowed and no longer covers this host.
  • The organization is suspended for billing or abuse; every resolution is refused while it is.
If none of those apply, the request is being refused on the host rather than by the policy, and it needs an operator. Open a ticket with the sandbox ID and the time.
For a host with a vault rule, the egress proxy presents a certificate from its own certificate authority (see How it works). Sandboxes trust it through /etc/ssl/certs/ca-certificates.crt, so curl, Python’s default SSL context and Node work. A client with its own certificate store does not: Python with certifi’s bundle passed explicitly, a Java keystore, or a pinned certificate.Point the client at the system bundle. In Python, drop cafile=certifi.where() or pass /etc/ssl/certs/ca-certificates.crt instead; for Java, import that file into the keystore the client uses. A certificate pinned for that host cannot match, so remove the pin.
Delete sandboxes you no longer need (anona-compute list, then anona-compute delete), or pause them: paused sandboxes do not count as running. For the vCPU limit, a smaller size also helps: on Free, size: "small" is 1 vCPU, and a sandbox created without a size from base is 2. The error’s limit_name, limit and current say which limit and how much is in use. See limits.
A plan limit that only you can clear: an API key, template or paused sandbox beyond your plan’s count, template storage full, a size above your plan’s largest, or on Free the month’s included vCPU-hours or egress used up (included_vcpu_hours, included_egress_gb; it resets on the first of the month). limit_name says which. Delete something you no longer need, or upgrade the plan. See limits.
Anona has set a maximum lifetime on your team; no plan has one. Use a timeout up to that maximum, or open a ticket to have it removed.
These E2B features are not available on Anona Compute; remove them from the create call. See Migrate from E2B.
Host names in allow_out need public egress off: add deny_out: ["0.0.0.0/0"]. Destinations in private or metadata ranges, and names that resolve into them, are always refused. Wildcard names such as *.example.com are not supported yet: name each host. See Networking.
The platform is not ready to serve the create yet: with sandbox_capacity_unavailable, every host is full and another is starting, which takes a few minutes. The CLI waits up to --capacity-wait (10 minutes). The stock E2B SDK retries for about a minute; the Anona SDK extras, from 0.3.0, wait up to 10 minutes (Python, TypeScript). If you call the API yourself, wait Retry-After seconds and retry while the response says Should-Retry: true.
The sandbox is paused, for example by its idle timer. Resume it with Sandbox.connect(id) or anona-compute resume <id>.
The certificate is older than 10 minutes, or was minted for another sandbox. Run anona-compute ssh <id> again, or refresh the block with anona-compute ssh <id> --print-config.
Tokens last at most 10 minutes. Answer the embed’s token-request message with a fresh token (Live view). After revoke_all, mint new tokens.
Port 49983 belongs to the Anona runtime (envd), so a template cannot expose it. Every FROM is pinned by digest at build time; a tag that cannot be resolved fails the build. See Templates.
These commands exist in the CLI ahead of their server side. See the CLI overview for what is coming.

Get more detail

  • anona-compute -v … prints verbose detail on stderr.
  • Every API error body has code, message and requestId, and often error_code. Include the requestId when you contact support.