> ## Documentation Index
> Fetch the complete documentation index at: https://docs.onecortex.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Limits

> Every limit in Onecortex in one table: organizations, builds, configuration, invoke rate limits, sessions, logs, names and the chat console.

These are the limits during the private beta. If one is in your way, email [support@onecortex.io](mailto:support@onecortex.io) with what you are building.

## Organizations

| Limit | Value |
| - | - |
| Organizations in the private beta, in total | 50 |
| Agents per organization | 20 |
| Members per organization | 10 |

## Builds and versions

| Limit | Value |
| - | - |
| Builds and rollbacks running at once, per organization | 3 |
| Builds running at once, per agent | 1 |
| Versions kept per agent, for rollback | 10 |
| Agent folder size | 50,000 files, 500 MB |
| Image size | 2048 MB |
| Smoke test: time to answer the health check | 30 seconds |
| Smoke test: time to answer its one invocation | 60 seconds |

A build that finds every slot taken fails with a message saying so. See [Builds](/deploy/builds).

## Configuration

| Limit | Value |
| - | - |
| Configuration entries per agent, secrets and variables together | 45 |
| Key format | Upper case letters, digits and `_`, not starting with a digit: `^[A-Z_][A-Z0-9_]*$` |
| Reserved keys | `PORT`, and platform prefixes such as `ONECORTEX_` and `OTEL_` |

See [Configuration and secrets](/build/configuration).

## Invoke

| Limit | Value |
| - | - |
| Requests per organization | 60 a minute, with a burst of 20 |
| Requests per agent | 30 a minute, with a burst of 30 |
| Streams open at once, per organization | 10 |
| Request body | 1 MB |
| `metadata` | 4 KB, as JSON |
| `sessionId` | 1 to 256 characters |
| A tool result's `output` | 64 KB, cut beyond that |
| Stream heartbeat, while your agent is silent | every 15 seconds |

Both request limits apply to every call, and the organization limit is counted first, so calls to one agent also spend the organization's budget. The limits refill continuously: after a burst, calls go through again at the sustained rate. Over a limit, the answer is `429` `rate_limited` with a `Retry-After` header in whole seconds. Every call that reaches your agent has `X-RateLimit-Remaining`, the requests left in the organization's burst.

A retry loop should wait for `Retry-After`, never retry at once. See [Timeouts and retries](/production/timeouts-and-retries).

## Sessions

| Limit | Value |
| - | - |
| Idle time before a session's instance stops | 15 minutes |
| Longest a session's instance runs | 1 hour |

After either, the next call with that `sessionId` starts a fresh instance. See [Sessions](/build/sessions).

## Logs

| Limit | Value |
| - | - |
| Log retention | 14 days |

## Names

| Limit | Value |
| - | - |
| Agent name | 1 to 60 characters |
| Agent slug | 2 to 40 characters: lower case letters, digits and single hyphens |
| Agent description | 280 characters |
| API key name | 1 to 60 characters |

## Chat console

| Limit | Value |
| - | - |
| Chats kept per agent | 50 |
| One message | 200,000 characters |
| Chat title | 60 characters |
| Events kept per reply | 200 |

See [The chat console](/observe/chat).