Add the bashOutputMaxChars setting to your settings.json file and set it to a figure as high as 128000 — that is the whole trick to increase Claude Code bash output limit caps, and it works from Claude Code v2.1.261 onwards, according to the official Claude Code changelog on GitHub (Releasebot's Anthropic tracker dates that release to 5 September 2026). The same release added a sibling setting, taskOutputMaxChars, which does the identical job for background-task output. Between the two of them, you control exactly how much command output Claude actually reads inline before the harness gives up and saves the rest to a file.
📺 Watch: Claude Code Just Got a HUGE Customization Upgrade
🔥 Get the Agent OS as a free bonus: AI Profit Boardroom members get the full Agent OS zip, prompt libraries, daily tutorials and weekly live coaching calls. → Get inside · Want AI SEO help 1-on-1? Book a free SEO strategy session →
This one matters more than it looks. Anyone who runs real build tools, test suites or scrapers through Claude Code has watched a long command finish, only for Claude to receive a truncated slice of the output with a note that the rest was written to disk. Claude then has to spend another tool call reading the file back — or worse, it reasons from the truncated slice and misses the failing test at the bottom of the log. Raising the bash output limit removes that round trip for the workloads where it hurts most.
What the Bash Output Limit in Claude Code Actually Is
Claude Code does not pipe unlimited terminal output into the model. Every Bash command Claude runs has its output captured by the harness, and only a capped number of characters is delivered inline into the conversation; anything beyond the cap is saved to a file that Claude can read later if it decides it needs to. The cap exists for a sensible reason — context is finite and expensive, and a single careless command can dump megabytes of logs — but the default is a compromise, and compromises fit nobody perfectly.
The changelog entry for v2.1.261 describes the new controls plainly: bashOutputMaxChars and taskOutputMaxChars were added to raise how much command and background-task output Claude receives inline before it is saved to a file, up to 128K characters. In other words, the ceiling is 128000 characters per output, and you choose where below that ceiling your own limit sits. That is roughly 30,000 tokens of raw text at the upper end, so treat the maximum as a scalpel for specific jobs rather than a default for everything.
How to Increase Claude Code Bash Output Limit Settings
The change is a two-minute edit. Claude Code reads its configuration from settings.json — your user-level file lives in the .claude folder in your home directory, and a project-level file in the project's own .claude folder overrides it. Here is the process:
- Update Claude Code first. The settings only exist in v2.1.261 or newer, so run your usual update path and confirm the version before editing anything.
- Open your settings.json. Use the user-level file if you want the new bash output limit everywhere, or the project-level file if one repository produces unusually heavy logs and the rest of your work does not.
- Add bashOutputMaxChars with your chosen number. A value like 60000 is a meaningful raise for most people; 128000 is the maximum the harness will honour.
- Add taskOutputMaxChars alongside it if you also run background tasks — long test runs, dev servers, watchers — and want their output held to the same standard.
- Restart Claude Code. Settings are read at startup, so restart the session and run a deliberately chatty command to confirm more of the output arrives inline.
That is genuinely all there is to it. No flags, no environment variables, no plugins — just two settings keys that did not exist before September 2026.
If you want Claude Code workflows that are already tuned like this out of the box — settings, prompt libraries and the full Agent OS zip — check out the AI Profit Boardroom. Prefer to talk it through 1-on-1 first? You can book a free SEO strategy session and map your setup with Julian directly.
bashOutputMaxChars vs taskOutputMaxChars
The two settings are twins with different jobs, and it is worth being precise about which one your problem actually needs before you raise either bash output limit.
| Setting | What it caps | Raise it when |
|---|---|---|
| bashOutputMaxChars | Inline output from Bash commands Claude runs in the foreground | Test suites, builds, linters and scripts whose important lines land at the end of long output |
| taskOutputMaxChars | Inline output from background tasks before it is saved to a file | Long-running background jobs — servers, watchers, batch scripts — that Claude checks on mid-run |
Both share the 128K-character ceiling, and both trade context budget for completeness. If you only ever notice truncation on ordinary commands, leave taskOutputMaxChars alone and adjust the bash side only.
When a Higher Bash Output Limit Helps — and When It Hurts
Raising the limit is not free. Every extra character Claude receives inline is context spent, and context is the resource that quietly drives your bill and your session length. The cases where a higher bash output limit clearly pays for itself look like this: a test runner that prints hundreds of passing lines before the one failure that matters; a build whose actual error sits below thousands of warnings; a data script whose summary table lands at the very bottom. In those situations, truncation forces Claude into an extra read-the-file step every single time, and the follow-up read often pulls in the whole file anyway — so the cap saved you nothing and cost you a round trip.
The cases where it hurts are just as predictable. If your commands routinely produce noise — verbose package managers, chatty install scripts, debug-level logging — a raised limit means Claude wades through all of it on every call, your token usage climbs, and useful context gets crowded out. The smarter play there is the opposite discipline: keep the limit modest and cut the noise at source. The funnel guide on how to reduce Claude Code token usage covers that side of the trade in detail — tool output flooding is listed there as the number one token leak, and raising your output cap amplifies exactly that leak if you apply it indiscriminately.
A sensible middle path: set a project-level bashOutputMaxChars only in the repositories where truncation genuinely bites, and keep your user-level default conservative. Settings hygiene like this is a core habit inside Agent OS, which treats configuration as part of the operating system you build around your agents rather than an afterthought. And if you are weighing up which model to point all this freshly enlarged output at, the Goldie Bench write-up covers how the current agent brains compare in hands-on tests.
📺 Watch: Free Claude Code Is CRAZY GOOD!
What Else Shipped in Claude Code v2.1.261
The bash output limit settings headline the release, but v2.1.261 carried several related quality-of-life changes worth knowing about, per the same changelog:
- An Organisation policy line in /status and claude doctor — it now says why your organisation's policy could not be loaded, such as a proxy not passing the endpoint through.
- A skill-doctor command — it shows which loaded skills go unused and what they cost in context, so you can prune them. If you are raising output limits with one hand, this is the tool for clawing context back with the other.
- A new flag for reading subagent system prompts from a file — aimed at prompts too large to pass on the command line.
- Input fixes — typed or pasted characters occasionally landing out of order during fast input were fixed in the same release.
If you are still getting oriented in the tool, the broader guides on learning Claude Code and running Claude Code free are the right starting points before you begin fine-tuning limits.
Common Questions About the Claude Code Output Limit
What is the maximum value?
128000 characters, per the changelog wording — the settings raise inline output up to 128K characters. Values above that are pointless; the harness will not deliver more.
Does raising the limit increase token usage?
Yes, whenever commands actually produce more output than the old cap. Characters delivered inline are tokenised and billed like any other context. Raise it where truncation costs you round trips, not everywhere.
Where does the overflow go?
Same place as before: output beyond your configured cap is saved to a file, and Claude can read that file in a follow-up tool call if it needs the rest.
Do I need this if I use MCP servers or restricted setups?
The settings are independent of your tool surface. They sit happily alongside managed MCP servers, restricted-access configurations and rendering tweaks like no-flicker mode — none of those change how command output is capped.
Does this affect subagents and orchestration?
Background-task output is exactly what taskOutputMaxChars governs, which matters once you start running heavier multi-agent orchestration in Claude Code — the more agents working in parallel, the more task output competes for inline delivery.
Verdict: A Small Setting With Outsized Leverage
The ability to increase Claude Code bash output limit values is the kind of unglamorous change that quietly improves every heavy session you run. One settings key, a number under 128000, a restart — and the failing test at the bottom of a long log reaches Claude on the first pass instead of the second. Pair it with deliberate token hygiene and it is pure upside; apply it blindly and you have simply bought yourself a bigger firehose. Configure it per project, keep the default lean, and let the noisy repositories earn their raise.
If you want the full stack — the Agent OS zip, tuned Claude Code settings, prompt libraries, daily tutorials and weekly live coaching — check out the AI Profit Boardroom → get the complete setup here. And if you would rather have your AI SEO pipeline looked at personally first, book a free SEO strategy session and bring your messiest log file with you.
