# Release notes

## 0.1.0

Initial release of `llsh`, a static Linux terminal LLM agent.

### Included

* Interactive and one-shot CLI modes with OpenAI-compatible, Anthropic, and
  Ollama providers.
* Built-in file, shell, embedding, one-shot LLM, skill, memory-search, and
  context-compaction tools.
* `.agents/` support for model presets, system prompts, skills, sub-agents,
  tasks, memories, and MCP configuration.
* Session save/resume, transcript export, context management, and goal loops.
* Bundled third-party sources, host GCC build support, and a static
  x86_64 Linux executable.

### Security and compatibility notes

* MCP servers from `.agents/mcp.json` require explicit opt-in with
  `--allow-mcp`.
* MCP tools use qualified names of the form
  `mcp__<server-name-length>_<server-name>__<tool-name>`.
* Model tool execution enforces the active profile's tool allowlist and the
  shell policy gate.
* The executable targets x86_64 Linux and uses the host glibc ABI. Glibc name
  service lookups may use NSS shared libraries at runtime; HTTPS requests also
  require a usable system CA certificate bundle.

### Verification

A clean vendor-cache rebuild completed with `make test`. `make asan` also
passed in the development environment.