GitHub issue #59401 in zed-industries/zed, opened 2026-06-16 by chrrity, records a report that large tool outputs from grep and read_file are sent without truncation in Zed agent threads, that accumulated raw tool outputs exceeded 1M tokens by the fourth or fifth user message in the reported thread, that the thread JSON was 13.3 MB uncompressed with 99.7 percent raw tool execution output, and that a single grep call returned 5.2 million characters which gets stored and re-sent on every subsequent request. yeskunall replied in the thread that a lot of what was identified had actually been addressed since version 1.6.3, and the author clarified in a comment the same day that the source code analysis was on a stale local checkout, while stating the crash itself is real and the remaining gap is grep and read_file lacking output truncation.
Verbatim Excerpts from Cited Records
“2. Have a conversation spanning 3+ user turns with multiple tool calls per turn (grep, read_file, terminal) 3. By the 4th–5th user message, the accumulated raw tool outputs exceed 1M tokens”
·
zed-industries/zed (issue 59401 opening body, Reproduction steps section)
·
2026-06-16T03:20:52Z
·GitHub Issue ↗
“The thread JSON for this conversation is **13.3 MB uncompressed** — **99.7%** of which is raw tool execution output, not user/agent text”
·
zed-industries/zed (issue 59401 opening body, Observed Behavior section)
·
2026-06-16T03:20:52Z
·GitHub Issue ↗
“There is no distinction between "current turn tool results" (needed for the ongoing tool-use loop) and "historical tool results from previous turns" (already digested by the agent's thinking blocks).”
·
zed-industries/zed (issue 59401 opening body, Root Cause section)
·
2026-06-16T03:20:52Z
·GitHub Issue ↗
“A single grep call can return **5.2 million characters** of output that gets stored and re-sent on every subsequent request.”
·
zed-industries/zed (issue 59401 opening body, Root Cause section)
·
2026-06-16T03:20:52Z
·GitHub Issue ↗
Counter-record: Issue open; API returned 23 comments, all 23 read. yeskunall replied 2026-06-16 that automatic compaction, cumulative token usage tracking, a terminal tool output cap, and MaxTokens handling had been addressed since 1.6.3, and that the remaining gap is non-terminal tool outputs not being truncated at assembly. The author clarified the same day that the source code analysis was on a stale local checkout and some listed items may already be fixed, while stating the crash itself is real and the remaining gap is grep and read_file lacking output truncation. Other users corroborated the report on later versions through 1.19.2 and other models, including ChatGPT subscription and OpenRouter. No retraction of the report appears in the thread. The claim documents only the original report and the thread discussion, not a verified current state.
Falsification probe: If issue 59401 is deleted without archive, if the author edits the opening body so the quoted sentences no longer match verbatim, or if the author posts a retraction of the report, this claim drops to Withdrawn.
Audit verification: Verified 100% continuous verbatim substring against official GitHub REST API by
scripts/verify_claim_verbatim.py.