Skip to content

Huge memory occupation by sequentialthinking MCP #2912

Closed

Description

@MekayelAnik

Describe the bug

After using sequentialthinking for a few hours, the RAM gets hogged up by this MCP server as much as 10GB+ RAM or more (I had this much RAM available, so..). But I doubt it is the Planning, because 10+ GB of Text is a huge amount of text, I DON'T THINK it was the TEXT. I strongly think there is a memory leak.

To Reproduce

  1. Use this MCP for a few (6-8) hours for frequent and long planning

Expected behavior
The memory should be freed after a job is done, or the memory should be offloaded to a disk space after a certain period of time. This should also help with persistent planning (Maybe, add a new tool for this).

Additional context
I have deployed this MCP as a distributed, standalone Docker container.

I love this MCP, a really great project. MANY MANY THANKS to the modelcontextprotocol Team
Regards

Activity

  1. added
    server-memoryReference implementation for the Memory MCP server - src/memory
    on Oct 29, 2025
  2. added
    server-sequentialthinkingReference implementation for the Sequential Thinking MCP server - src/sequentialthinking
    and removed
    server-memoryReference implementation for the Memory MCP server - src/memory
    on Dec 8, 2025
  3. added a commit that references this issue on Feb 11, 2026
    ceb2381
  4. cliffhall commented on Oct 19, 2026

    @cliffhall
    Member

    Triage: Priority Low (total 5)

    • Severity 2: The sequential-thinking server keeps every thought and branch for the life of a connection with no cap or reset, so memory grows without bound on a long-lived connection; the 10 GB reported likely also involves the third-party container wrapper, which could not be checked.
    • Urgency 2: wanted eventually
    • Bonuses: +1 bug label

    Still present on v2 (checked by reading the current code).

    Board: Servers V2 (#43), Status Incoming, awaiting maintainer review (no milestone yet). Part of the backlog triage in #4875.

  5. cliffhall commented on Oct 20, 2026

    @cliffhall
    Member

    Prior work, from the PR backlog triage (#4877). These outside pull requests addressed this issue against the pre-refactor code:

    They are being closed as part of the backlog triage in #4875, since the repository now takes issues rather than outside PRs and their diffs predate the v2 refactor. They stay readable and serve as prototypes for the fix here. Thank you to their authors.

  6. added this to the v2.3.0 milestone on Oct 25, 2026
  7. self-assigned this
    on Oct 28, 2026
  8. cliffhall commented on Oct 29, 2026

    @cliffhall
    Member

    Fixed by #5397, merged into v2/main. Ships in v2.3.0.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

bugSomething isn't workingserver-sequentialthinkingReference implementation for the Sequential Thinking MCP server - src/sequentialthinkingv2

Type

No type

Projects

No projects

    Milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions