HoldKit

Is HoldKit MCP server safe?

• None of the hidden-instruction wordings we know were found in its tool descriptions.
• Its error messages showed no secrets or internal details.
• 3 of its 14 tools are labeled read-only; the others may change or send things.
• Overall it scored 93/100 on 31 Oct 2026.

A test can only see what the server shows from outside, not what it does with your data behind the scenes. Connect servers from makers you trust.

What it can do

14 tools, in the server's own words.

Test results

Safe mode: only tools labeled read-only were run. How the test works

• !
  Every tool is described

  16 warnings

  • !inbox.poll: Inputs with no description: inbox_id, cursor, limit
  • !inbox.ack: Inputs with no description: inbox_id, event_ids
  • !inbox.wait: Inputs with no description: inbox_id, cursor, block_ms
  • !timer.set: Inputs with no description: delay_s, inbox_id, url, payload
  • !timer.cancel: Description is very short (15 characters)
  • !timer.cancel: Inputs with no description: timer_id
  • !timer.cancel: Description only repeats the tool name, so it adds nothing an AI could use: "Cancel a timer."
  • !lock.acquire: Inputs with no description: name, ttl_s, owner
  • and 8 more
  note
• !
  Follows MCP best practices (not scored yet)

  2 tips (not scored)

  • !11 tools not labeled read-only don't say whether running them twice is safe (idempotentHint): keys.create, inbox.create, inbox.ack, timer.set, timer.cancel, lock.acquire and 5 more
  • !14 tools don't say whether they reach outside services like the web (openWorldHint): keys.create, inbox.create, inbox.poll, inbox.ack, inbox.wait, timer.set and 8 more
  note
• ✓
  ConnectsConnected in 198 ms
  pass
• ✓
  Ready for the newest MCP versionUses the latest MCP version (2026-07-28)
  pass
• ✓
  Lists its toolsFound 14 tools
  pass
• ✓
  Tools are labeled safe or dangerousEvery tool has a safety label
  pass
• ✓
  No known hidden instructionsNone of the hidden-instruction wordings we know were found in tool descriptions. New wordings can get past this check.
  pass
• ✓
  Tools work with good inputs

  All good

  • ✓inbox.poll: Answered with an error, probably because of our sample input: {"ok":false,"reason":"caller_key_required"}
  • ✓inbox.wait: Answered with an error, probably because of our sample input: {"ok":false,"reason":"caller_key_required"}
  • ✓kv.get: Answered with an error, probably because of our sample input: {"ok":false,"reason":"caller_key_required"}
  • –keys.create: Not run: labeled as changing data
  • –inbox.create: Not run: labeled as changing data
  • –inbox.ack: Not run: labeled dangerous
  • –timer.set: Not run: labeled as changing data
  • –timer.cancel: Not run: labeled dangerous
  • and 6 more
  pass
• ✓
  Bad inputs give a clean error

  All good

  • ✓inbox.poll: Refused missing required input "inbox_id" with a clear error
  • ✓inbox.poll: Refused wrong type for "inbox_id" with a clear error
  • ✓inbox.wait: Refused missing required input "inbox_id" with a clear error
  • ✓inbox.wait: Refused wrong type for "inbox_id" with a clear error
  • ✓kv.get: Refused missing required input "key" with a clear error
  • ✓kv.get: Refused wrong type for "key" with a clear error
  pass
• ✓
  Fast enough

  All good

  • ✓Connecting took 198 ms
  • ✓Listing tools took 236 ms
  • ✓Average tool answer: 189 ms
  • ✓inbox.wait: Slowest tool: inbox.wait (211 ms)
  pass
• ✓
  No secrets leakedChecked 9 error messages: no secrets or internal details
  pass
• –
  Tools work togetherNo safe tool returned data to start a chain from, so chains were not tested (not counted either way)
  not tested

Test it yourself

Free and open source. Same checks, same score.

Show this score on GitHub

Free for every server in the directory. Paste this into your README: the badge shows the score from our latest test.

Score over time

• 31 Oct 202693

One test so far. New tests will show here.

Reviews

Coming soon: reviews from people who use HoldKit, signed in with GitHub, one per person. Reviews never change the score.