agents · review · first seen Oct 24, 09:32 UTC

CrewAI releases 1.15.30 with AGENTS.md fix

CrewAI shipped 1.15.30 with an AGENTS.md fix for login during evaluation and a documentation update for v1.15.29. The reason to check it today is narrow: if your evaluation workflow uses AGENTS.md, its login instructions now have a stated start point and stopping point.

L0 · primary1 source2 claims checked3 min read

What shipped

The bug-fix entry says AGENTS.md starts the login process while evaluating and stops after three rounds. That is the behavior the 1.15.30 release note identifies. I read this as a fix to the AGENTS.md instructions, not as evidence that the evaluation engine, authentication service, or a separate login implementation changed. That distinction matters before treating another login problem as resolved. [1]

The release also updates the snapshot and changelog for v1.15.29. It lists this as documentation, separate from the AGENTS.md bug fix. The updated documents concern v1.15.29; the release carrying the update is 1.15.30. I would keep those version references distinct in review notes. The excerpt does not say what either document said before the update. [1]

These are the two shipped items established by the supplied release excerpt. It gives one specific login boundary: start while evaluating, then stop after three rounds. It gives no migration steps, other options, or measured effects. The useful detail for a builder is that instruction boundary, not a general claim about evaluation quality. [1][2]

Why it matters if you build

If your evaluation workflow uses AGENTS.md, you can compare its login instructions with the released wording. Check whether they place the start during evaluation and the stop after three rounds. That is a review of what the instructions say. The release note alone cannot tell you whether a login attempt in your workflow will succeed. [1]

The three-round stopping point is a limit worth keeping visible during that review. Continuing the login process beyond it would not match the instruction described in the release note. The source does not define a round, describe what happens when login fails, or say whether an evaluation already in progress is affected. I would not derive a retry policy, reliability claim, or cost estimate from this line. [1]

The documentation update calls for a different check. If you rely on the v1.15.29 snapshot or changelog, review their updated wording before citing them. Nothing in this excerpt says that documentation update changed executable behavior. Keep it separate from the AGENTS.md fix when deciding what to test or explain to your team. [1][2]

How to try it today

1. Read the AGENTS.md bug-fix line in the 1.15.30 release note.
2. Inspect the AGENTS.md instructions used by your evaluation workflow for when login starts.
3. Check whether those instructions stop the login process after three rounds. Record any mismatch without assuming its cause.
4. Review the updated v1.15.29 snapshot and changelog if you rely on either document.
5. Decide whether your own evaluation instructions need attention; do not infer a broader login fix from this note.

What we checked

Every claim below was walked down the source ladder. The rung is how close the source is to the thing itself; the tense is the source's own.

1. shipped L0 · primary

   CrewAI 1.15.30 fixes AGENTS.md to start the login process while evaluating and stop after three rounds.

   per GitHub releases · crewAIInc/crewAI (L0)
2. shipped L0 · primary

   CrewAI 1.15.30 updates the snapshot and changelog for v1.15.29.

   per GitHub releases · crewAIInc/crewAI (L0)

What is not confirmed

The excerpt does not establish whether login succeeds in a running evaluation, how a round is defined, whether executable code changed, or whether other workflows use the same instructions. It also does not describe the contents of the v1.15.29 snapshot and changelog update. Those remain questions to check, not outcomes established by the release note. [1][2]

Who should act, who should wait

Act now if you maintain an evaluation workflow that uses AGENTS.md: inspect the instruction and compare its start and stop points with the shipped note. Review the v1.15.29 documentation too if you cite it. Wait before changing runtime retry policy, promising successful logins, or attributing other evaluation outcomes to 1.15.30. The supplied source supports the instruction fix and documentation update, but not those broader conclusions. [1][2]

Sources

• 1.15.30 — GitHub releases · crewAIInc/crewAI L0 · primary