Skip to content

AgentiLoopGo 0.0.5

Latest
Latest

Choose a tag to compare

View all tags
@github-actions github-actions released this 01 Oct 04:31
v0.0.5
1ae4204

🎉 Release — 0.0.5

AgentiLoop 0.0.5 release screenshot: Agent! for macOS running the Rust and Go CLIs side by side in the full-screen TUI

🎉 Release

AgentiLoopGo 0.0.5 is built from main and tagged v0.0.5. The theme of this release is a bigger toolbox: the agent can now search your project, fetch web pages and keep a checklist, it reads your AGENTS.md, you can undo its file changes, and you can script it with --json, custom slash commands and tool allow/deny rules. The Rust version, AgentiLoop CLI 0.0.5, has the same features. The macOS binaries are signed with Developer ID and notarized.

Testers welcome — see What to test below.

✨ New

  • Project search tools: read-only glob and grep tools skip .git, node_modules, target and binary files, honor .gitignore (nested files, negation, anchoring, directory-only), and grep can show 0-5 lines of context around each match (9b40269, 070ed6d, a09a41f).
  • AGENTS.md / CLAUDE.md: project and ~/.agentiloop instruction files are loaded into the system prompt, and @path lines import other files (nested, cycle-safe) (2671d2a, f5fc4d4).
  • web_fetch tool: fetch an http(s) page as text, with size and character caps; it asks before running (c2c81a5).
  • todo_write tool and /todos: the agent keeps a checklist for multi-step work (018805f).
  • /undo: file changes from write_file, edit_file and apply_patch are journaled per prompt and can be reverted (7c7427a).
  • /init creates a starter AGENTS.md with detected build and test commands (b64adbc).
  • /export saves the conversation as Markdown (991208a).
  • /usage shows token totals since start and how full the context is (cf0f993).
  • /diff shows the git status and diff of the working tree (d0053c1).
  • Custom slash commands: put Markdown files in .agentiloop/commands/*.md ($ARGUMENTS, $1..$9), list them with /commands (bbeb156, 9534157).
  • MCP prompts: server prompts become /mcp__<server>__<prompt> commands (aa2fc82).
  • --json: a one-shot run prints the answer as a single JSON object (result, is_error, session_id, provider, model, usage) (740831d).
  • --allow-tool / --deny-tool: permission rules by tool name or mcp_* prefix; deny beats allow and --yes (622b8a7).
  • --append-system-prompt: add text to the system prompt for one run (5266224).
  • Piped input: a lone - in a one-shot prompt is replaced by what is piped on stdin, e.g. git diff | agentiloop "review this" - (493ac23).

🔧 Changes

  • README: refreshed --help block and docs for all of the above in all 8 languages (70aaa0a).

📦 Downloads

Platform File
macOS Apple Silicon agentiloop-macos-arm64.tar.gz (signed + notarized)
macOS Intel agentiloop-macos-x86_64.tar.gz (signed + notarized)
Linux x86_64 agentiloop-linux-x86_64.tar.gz (glibc 2.35+)
Linux ARM64 agentiloop-linux-arm64.tar.gz (glibc 2.35+)
Windows x86_64 agentiloop-windows-x86_64.zip

Unpack the download and put agentiloop (agentiloop.exe on Windows) on your PATH, then run agentiloop — the wizard takes it from there.

🧪 What to test (testers welcome!)

  • agentiloop --version prints 0.0.5.
  • Ask the agent to find something in your project: it should use glob/grep and skip ignored files. Ask it to read a web page (web_fetch asks first).
  • Create an AGENTS.md (or run /init), start a new session and check the agent follows it.
  • Let the agent edit a file, then run /undo — the file should be restored. Try /diff, /usage and /export.
  • Add .agentiloop/commands/review.md containing Review $1, then run /review main.rs; /commands should list it.
  • git diff | agentiloop "review this" - and agentiloop --json "say hi" print what you expect.
  • agentiloop --deny-tool bash "run ls" — the model should be told bash was denied.
  • Report problems with your OS, provider/model and the command you ran. Do not include API keys.

🔢 Version

Full Changelog: v0.0.4...v0.0.5

Assets 7
Loading