Skip to content

Plans & Limits

LokAI plan enforcement is split into:

  • Capabilities: boolean feature flags such as workspace tokens or tenant customizations
  • Quotas: numeric thresholds such as projects, glossary entries, or PATs per user

Capabilities are resolved from the plan plus workspace overrides. Quotas are enforced at mutation call sites and may reject requests before data is created.

Capability Free Pro Team Enterprise
custom_roles Off Off On On
teams Off On On On
workspace_tokens Off On On On
tenant_customizations Off Off On On
audit_log_export Off Off On On
llm_translation_review Off On On On

custom_roles and audit_log_export are reserved for features that have not shipped yet: you cannot create custom roles or export the access log today. The plan values above apply once they ship.

Quota key Free Pro Team Enterprise
projects 1 5 25 unlimited
keys_per_project 250 5,000 50,000 unlimited
keys_per_workspace 250 15,000 150,000 unlimited
base_languages_per_project 5 15 25 unlimited
base_languages_per_workspace 5 15 50 unlimited
members 3 10 50 unlimited
pats_per_user 3 10 25 50
wats_per_workspace 0 5 25 unlimited
tenants 0 0 10 unlimited
rules_per_tenant 0 0 50 unlimited
glossary_entries 25 100 250 unlimited
llm_calls_monthly 1,000 + 10% grace 25,000 + 10% grace 250,000 + 10% grace contract-defined + 10% grace
  • llm_calls_monthly is currently the only shipped quota using hard_grace behavior.
  • The 10% grace is computed on the plan value only. Extra calls you buy raise your limit but do not add grace.
  • Enterprise still caps pats_per_user at 50; the rest of the matrix is contract-defined or unlimited where shown.
  • Base-language quotas count base language groups, not every regional variant independently.

These operational limits apply to every plan, independently of plan quotas. One MiB is 1,048,576 bytes.

Input Limit
Default HTTP request body, including key bulk create and tRPC batches 1 MiB
Glossary import body over REST or standalone tRPC 8 MiB
Glossary content, measured as UTF-8 bytes after decoding the request 1 MiB
Glossary entries per import 1,000
Keys per bulk-create request 100

The larger import body allowance accommodates JSON escaping of the embedded file content. All tRPC batches retain the default 1 MiB body limit, including batches that contain glossary imports. Send larger glossary imports as standalone calls.

Exceeding a body or content byte limit returns HTTP 413 PAYLOAD_TOO_LARGE. Exceeding the glossary entry count returns 400 BAD_REQUEST before any entries are written. JSON and YAML count elements of entries; CSV counts grouped entries, not translation rows.

See Import and export for splitting files and handling partial failures.

These tables describe what is seeded today, not every future feature in the data model.

Still deferred:

  • custom-role CRUD
  • branch-aware authorization
  • per-language coordinator UX preset
  • REST API — includes the capabilities endpoint used to inspect live quota state
  • Architecture — workspace, project, token, and grant concepts