Skip to content

docs(mcp): document new session reclamation behavior#13

Open
mgoldsborough wants to merge 1 commit into
mainfrom
docs/mcp-session-reclamation
Open

docs(mcp): document new session reclamation behavior#13
mgoldsborough wants to merge 1 commit into
mainfrom
docs/mcp-session-reclamation

Conversation

@mgoldsborough
Copy link
Copy Markdown
Contributor

Summary

Companion docs update for nimblebrain#229.

MCP_MAX_SESSIONS is no longer a hard cap that returns 429 Too many active sessions — it's a memory-budget knob with LRU eviction. MCP_SESSION_TTL_SECONDS now also drives a local idle sweep on the serving pod (not just the session-store TTL).

Pages updated

  • config/environment.mdx — rewrites the MCP_MAX_SESSIONS row to reflect LRU and points operators at the [mcp] evicting transport reason=pressure log line as the signal to raise the cap. Tightens MCP_SESSION_TTL_SECONDS to mention the local sweep alongside the session-store TTL.
  • api/mcp-endpoint.mdx — same change in the Sessions table; updates the "Session limits" security bullet so it doesn't claim 429 prevents runaway clients (LRU does, and more correctly).

Builds cleanly via npm run build. No content removed; no link targets changed.

Test plan

  • npm run build — 72 pages built, no errors
  • Visual spot-check after deploy

`MCP_MAX_SESSIONS` is now a soft cap: a new initialize at capacity
evicts the least-recently-used session via LRU instead of returning
`429 Too many active sessions`. `MCP_SESSION_TTL_SECONDS` now drives a
local sweep on the serving pod in addition to the session-store TTL.

Updates the env-var reference and the `/mcp` endpoint docs to match,
and replaces the "session limits prevent exhaustion" line with one that
notes runaway clients can no longer deny service to others under LRU.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant