AgentStack
Browse Sign in
Browse Why AgentStack Sell Docs
Sign in
MCP unreviewed MIT Self-run

Mish

mcp-aetherwing-io-mish · by aetherwing-io

The LLM-native shell — five MCP tools for structured output, error diagnostics, and process control

No reviews yet
0 installs
12 views
0.0% view→install

Install

$ agentstack add mcp-aetherwing-io-mish

Open-source listing, not yet scanned by AgentStack. Follow the source repository for install instructions.

Security review

⚠ Flagged

1 finding(s); flagged for manual review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures
  • high Pipes remote content directly into a shell (remote code execution).

What it can access

  • Network access Used
  • Filesystem access No
  • Shell / process execution No
  • Environment & secrets No
  • Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

View the full security report →

Reliability & compatibility

Not yet reviewed
0 installs to date
no reviews yet
5mo ago

Declared compatibility

Claude CodeClaude DesktopCursorWindsurf

Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.

Preview Execution monitoring

We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.

How agent discovery & health will work →
Are you the author of Mish? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

Five MCP tools between your agent and the OS. Structured output · error diagnostics · process control · one binary

ohitsmish.com · Releases · Architecture · Benchmarks


Before:  agent → Bash("cargo test")        → 1,319 lines raw output
After:   agent → sh_run("cargo test")      → 201 lines (6.6× reduction)

Before:  agent → Bash("cargo test") + watch → 1,319 lines raw output
After:   agent → sh_run watch="warning"     → 3 lines (440× reduction)

Before:  agent → Bash("cp missing dest/")  → "No such file" + 4 follow-up calls
After:   agent → sh_run("cp missing dest/") → error + path walk + permissions + nearest dirs

Install

# macOS (Apple Silicon + Intel) and Linux
curl -fsSL https://ohitsmish.com/install.sh | sh

# or from source
cargo install --git https://github.com/aetherwing-io/mish

Connect

# Claude Code
claude mcp add mish -- mish serve

# Cursor / Windsurf / any MCP client — add to your config:
"mish": { "command": "mish", "args": ["serve"] }

Restart your client. Five tools appear:

| Tool | What it does | |------|-------------| | shrun | Run a command. Output is squashed, categorized, and enriched with diagnostics on failure. | | shspawn | Start a background process. Wait-for-ready patterns, aliases, output spools. | | shinteract | Send input, read tail, signal, or kill a running process. | | shsession | Named PTY sessions with full lifecycle control. | | sh_help | Self-documenting reference card. Your agent reads it, never needs a manual. |

What it does

mish sits between the shell and its caller — whether that's an LLM agent, a human, or both — and returns structured, context-efficient responses. It categorizes every command and applies the right handler. Pure heuristics, no LLM in the loop. Every response includes exit code, timing, and command category. On failure, mish pre-walks paths, checks permissions, and lists nearby files — before your agent has to ask.

Command routing

Every command is classified into one of six categories:

| Category | Commands | Behavior | |----------|----------|----------| | Condense | npm, cargo, docker, make, pytest | PTY capture → ANSI strip → dedup → Oreo truncation | | Narrate | cp, mv, mkdir, rm, chmod | Inspect → execute → narrate what happened | | Passthrough | cat, grep, ls, jq, diff | Output verbatim + metadata | | Structured | git status, docker ps | Machine-readable parse | | Interactive | vim, htop, psql, node REPL | Transparent passthrough with raw mode detection | | Dangerous | rm -rf, force push, reset --hard | Warn before executing |

Benchmarks

Same commands, same machine, same session. mish vs bare shell on a real Rust codebase.

| Scenario | Reduction | How | |----------|-----------|-----| | Full test suite | 6.6× | Dedup + Oreo truncation — head and tail kept, repetitive middle dropped | | Test suite + watch | 440× | Only regex-matched lines return | | Failed command | 5→1 round trips | Path walks, permissions, nearest dirs — pre-fetched on failure |

Full benchmark data: [SHOWCASE.md](docs/SHOWCASE.md)

How it works

mish is one binary with two interfaces:

MCP server (mish serve) — a process supervisor over JSON-RPC with ambient process state on every response, watch patterns for regex filtering, and operator handoff for auth/MFA.

CLI proxy (mish ) — wraps individual commands with category-aware structured output. Works standalone or with any LLM tool.

Both share the same core: category router → squasher pipeline (VTE parse, progress removal, dedup, Oreo truncation) → error enrichment → grammar system (TOML tool grammars with dialect support).

See [ARCHITECTURE.md](docs/ARCHITECTURE.md) for the full execution model.

Status

Active development. 1,400+ tests across unit, integration, CLI, grammar, fixture, and MCP layers. MCP server is live and battle-tested in daily use with Claude Code. macOS and Linux.

License

[MIT](LICENSE)

Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet, be the first.

Versions

  • v0.1.0 Imported from the upstream source.