Elizon Docs

Error Codes

Error codes returned by the Elizon Public API.

Every error response is JSON. Most follow { "error": "<code>", "message": "<human-readable>" }; validation failures use { "errors": [{ "path", "message" }] } instead — see the table below for which shape applies where.

HTTP statuserrorWhen it occursSuggested action
400validation error ({ "errors": [...] })Request body or query params fail schema validation (missing required field, wrong type, out-of-range value)Fix the field(s) named in errors[].path; each entry's message explains what failed.
400invalid_tool_contexttool_context was provided but isn't valid JSONSend a JSON-encoded string, or omit the field.
401unauthorizedX-API-Key header missing, malformed, invalid, expired, or revokedCheck the header is present and correctly formatted (see Authentication). Elizon doesn't distinguish these cases in the response — if a previously working key suddenly fails, check whether it was revoked or has an expiresAt in the past.
403forbiddenA project-scoped key tried to access an agent or job that exists but belongs to a different projectUse the correct project-scoped key, or an org-scoped key.
404agent_not_foundNo agent with that ID exists in your organizationCheck the agentId — this is also returned (not 403) for agents in another organization entirely.
404job_not_foundNo job with that ID exists for your organization/projectCheck the jobId. Also returned for jobs belonging to another tenant — never 403 — so a job ID alone never confirms another tenant's job exists.
404run_not_foundCancelling a run (DELETE .../runs/:runId) with an unknown runIdCheck the runId returned by /run or the RunStarted streaming event.
422project_not_foundproject_ref_id / projectId doesn't belong to your organizationUse a project ID that exists in your org, or omit it.
422llm_configuration_not_foundconfig references an llm_config_id that doesn't exist or isn't connectedCall GET /agents/llm-providers to list valid, connected configurations.
429rate_limit_exceededOrganization or per-key rate limit exceededBack off for retry_after seconds (see Rate Limiting).
502bad_gateway / upstream_error / enqueue_failedThe agent runtime is unreachable, errored mid-run, or a job couldn't be queuedTransient — safe to retry with backoff.
504execution_timeoutA synchronous /run call didn't finish within 30 secondsSwitch that call to /run-async and poll for the result.