Rate limits
Per-key, per-minute limits and what happens when you exceed them
Rate limits are enforced by the Goono Open API per API key, per minute, independently per bucket. The MCP server adds no limit of its own. There are no per-second or per-workspace limits.
| Bucket | Per minute | Tools |
|---|---|---|
default | 300 | All read and review tools (get_me, list_*, get_*, search) |
content | 60 | get_note_file · get_review_file · get_research_draft (when content is included) |
write | 60 | create_folder · rename_folder · delete_folder · move_note |
Limits may be tuned operationally. The current values are reported in response headers.
Response headers
Every successful Open API response carries the remaining quota.
| Header | Meaning |
|---|---|
X-RateLimit-Bucket | Bucket this request belongs to |
X-RateLimit-Limit | Per-minute limit of the bucket |
X-RateLimit-Remaining | Calls left in the current window |
X-RateLimit-Reset | When the window resets (epoch seconds) |
When the limit is exceeded
The Open API responds with 429, a body of { "code": "RATE_LIMITED", ... }, and a Retry-After header in seconds.
What the MCP client actually receives
The MCP server does not forward HTTP headers. The client (LLM) receives a single tool result with isError: true and text in this shape:
Goono API 429 /workspaces/... — [RATE_LIMITED] rate limit exceeded — 요청 한도를 초과했다. 잠시 후 재시도할 것. — (requestId: ...)The Retry-After value is not included. The LLM is only told to retry shortly.
If you get rate-limited
- Do not fetch files in parallel. The
contentbucket (60/min) runs out first. The server instructions already say "only the files you need, sequentially". - Use
includeContent: falsewhen scanning research drafts; it skips the body and does not touch thecontentbucket. - Narrow
list_noteswith filters.tags,extensions, andwrittenAfterreduce calls compared with fetching everything and filtering in the LLM. - Start
searchin preview mode. Omittingsectionsreturns up to 20 hits per section in one call; drill down by naming a single section and paging up to 50. - Several workspaces mean several keys. Limits are per key, so they do not affect each other.