Atlas glob prints "(Results are truncated: showing first 100 results. Consider using a more specific path or pattern.)" because GlobTool passes a hard limit of 100 to Ripgrep.glob and marks the result truncated when exactly that many files come back. The fix is to narrow the search so fewer than 100 files match: add a directory prefix or a stricter extension to the pattern, or set `path` to a subdirectory so the search root is smaller. If you are really looking for code rather than filenames, use grep with a content pattern, and if you need to cover a large tree, iterate directory by directory rather than globbing the whole repository at once.
Why does Atlas glob truncate at 100 results
Atlas glob truncates because GlobTool passes a hard limit of 100 to Ripgrep.glob and marks the result truncated when exactly that many files come back. The output then ends with "(Results are truncated: showing first 100 results. Consider using a more specific path or pattern.)" and the rest of the matches are not returned.
The cap is a signal about the query, not a limitation to route around. A pattern matching more than 100 files is almost never a pattern that answers a specific question. If you glob for `**/*.ts` in a real repository, you get 100 files that happen to sort first and no indication of whether the one you wanted is among them, which is worse than useless because it looks like an answer. The truncation notice exists to tell you that the list you are holding is a prefix, and that the correct response is a better query rather than a longer list. The tool is asking you to be more specific, and the message says so in plain words.
How to narrow an Atlas glob pattern
Narrow an Atlas glob pattern by adding a directory prefix or a stricter extension so fewer than 100 files match. A pattern like `**/*.ts` matches an entire codebase, while a pattern scoped to one package and one file kind matches a set small enough to actually read.
Two levers do most of the work. The directory prefix cuts the tree: constraining the pattern to a package, a module, or a feature folder removes everything outside it before matching begins. The stricter extension cuts the file kind: if you want configuration and not source, say so in the pattern rather than filtering by eye afterwards. Combining both usually takes a match set from hundreds down to a handful. When you still cannot get under 100, the pattern is probably not the right expression of what you want, and the question to ask is whether you are searching for a filename at all or for something inside the files.
How to shrink the Atlas glob search root with path
Set `path` to a subdirectory in Atlas glob so the search root is smaller. Glob starts walking from `path`, so scoping it to one package or one module cuts the candidate set before the pattern is even applied, which is the most direct way to get back under the 100 result limit.
Scoping the root and narrowing the pattern do similar work from different directions, and the choice between them is about what you know. If you know which directory the files live in, set `path` and keep the pattern loose, because the root is doing the filtering. If you know the file kind but not the location, keep the root wide and make the pattern strict. When you know both, use both. The one caveat: `path` must be a directory, since GlobTool stats the resolved search path and rejects a file, so scoping to a specific file is not the way to narrow a search.
Should I use grep instead of glob in Atlas
Use grep with a content pattern in Atlas when you are really looking for code rather than filenames. A glob that matches over 100 files usually means the filename was never the distinguishing feature: what you actually wanted was the file containing a particular function, key, or string, and grep finds that directly.
Hitting the 100 result cap is often a sign that the question was posed to the wrong tool. Globbing for every TypeScript file in a repository and then reading them looking for a handler is a slow path to an answer that grep returns in one call. Atlas searches code with hybrid semantic and keyword retrieval fused by reciprocal rank fusion, and Atlas indexes code by AST declarations using tree-sitter, not blind line windows, so content search returns real declarations rather than incidental line matches. Filenames are the right key when you know the name. Content is the right key when you know what the code does.
How to verify the Atlas glob truncation is resolved
Verify by re-running the narrowed Atlas glob call: a query matching fewer than 100 files returns its full result set with no "(Results are truncated: showing first 100 results...)" notice at the end. The absence of that line means the list you are looking at is complete.
Completeness is the property that matters, and it is why the truncation notice deserves attention rather than a shrug. A complete list lets you reason about what is not there, which is frequently the point: confirming that only three files import a module, or that a pattern appears in exactly one package. A truncated list supports no such conclusion. If the notice persists after narrowing, iterate directory by directory rather than globbing the whole repo, running one scoped call per subtree and combining the results yourself. That is slower per call and far faster overall than repeatedly re-running a query that keeps returning the same first 100 files.
How to fix it
- 01Narrow the pattern, for example add a directory prefix or a stricter extension so fewer than 100 files match.
- 02Set `path` to a subdirectory so the search root is smaller and the match set shrinks with it.
- 03Use grep with a content pattern when you are really looking for code, not filenames.
- 04Iterate directory by directory rather than globbing the whole repo in a single call.
- 05Re-run glob and confirm the truncation notice no longer appears at the end of the output.
Frequently asked questions
- why does Atlas glob only show 100 results
- GlobTool passes a hard limit of 100 to Ripgrep.glob and marks the result truncated when exactly that many files come back. The output ends with a notice suggesting a more specific path or pattern.
- how do I get more than 100 results from Atlas glob
- You narrow the query rather than raising the cap. Add a directory prefix or a stricter extension to the pattern, or set `path` to a subdirectory so fewer files match.
- is my file missing if Atlas glob truncated the results
- It may be. A truncated list is a prefix of the real match set, so the file you wanted can be outside the first 100 returned. Narrow the search before concluding the file does not exist.
- how do I make an Atlas glob pattern more specific
- Add a directory prefix so the pattern only matches inside one package or module, and tighten the extension so it only matches the file kind you care about. Combining both usually gets well under 100 matches.
- should I use grep or glob when my Atlas glob keeps truncating
- Use grep with a content pattern when you are really looking for code, not filenames. A glob matching hundreds of files usually means the filename was never the distinguishing feature.
- how do I glob a large repository in Atlas
- Iterate directory by directory rather than globbing the whole repo in one call. Scope `path` to each subtree in turn so no single call hits the 100 result limit.
- how do I know my Atlas glob results are complete
- The truncation notice is absent. A call that returns fewer than 100 files ends without "(Results are truncated: showing first 100 results...)", which means you are seeing every match.
Try Atlas in your terminal
The terminal-native AI coding agent. Free core, single binary.
Install AtlasRelated guides
Atlas with GPT-5.5 Pro: One Hard Question at a Time in 2026
GPT-5.5 Pro is OpenAI's April 2026 maximum-effort reasoning tier at $30 / $180 per Mtok on a 1,050,000 token window. Invoke it deliberately in Atlas, then switch back.
Debug a Single Failing Test with Atlas in 2026
How to debug one failing test with Atlas in 2026: run it in isolation with bash, walk the call graph with the lsp tool, and fix the code, not the assertion.
Atlas vs Factory AI: Terminal AI Coding Agents in 2026
Compare Atlas and Factory AI, two terminal AI coding agents for developers in 2026. Explore features, pricing, privacy, and workflow to choose the best fit.
Atlas with Llama 4 Scout (Ollama): A 10M-Token Window on Local Hardware in 2026
Run Atlas on Llama 4 Scout (Ollama): a 67GB 16-expert MoE with a 10M-token context and image input, free self-hosted. The memory math before you pull 67GB.
Atlas with StarCoder2 15B (Ollama): The Provenance Choice in 2026
StarCoder2 15B (Ollama) is BigCode's 9.1GB code model with a transparent training corpus and a 16K context, Free (self-hosted). Atlas setup and tradeoffs for 2026.
Atlas for Svelte in 2026
Adopt Atlas, the terminal-native AI coding agent, for Svelte and SvelteKit projects in 2026. Enhance development with intelligent code search, secure local embeddings, and guided code generation.
Atlas with GPT-5 Mini: Full 400K Context at One Fifth the Price in 2026
GPT-5 Mini in Atlas: $0.25 per Mtok input and $2 per Mtok output, 5x cheaper than GPT-5 on both sides, with no reduction to the 400K context window.
Atlas with Gemma 4 31B (Ollama): the Flagship Gemma 4 Tag in 2026
Gemma 4 31B (Ollama) is the largest Gemma 4 tag: 20GB weights, a 256K tokens (262,144) context, Free (self-hosted), plus a 31b-coding-mtp-bf16 variant. Atlas setup.