Package Exports

@trymesh/cli/dist/local-tools.js

This package does not declare an exports field, so the exports above have been automatically detected and optimized by JSPM instead. If any package subpath is missing, it is recommended to post an issue to the original package (@trymesh/cli) to support the "exports" field. If that is not possible, create a JSPM override to customize the exports field for this package.

Readme

@trymesh/cli

A terminal-first AI engineering agent. Zero-config, multi-model, verifiable.

Mesh runs in your terminal, indexes your codebase via AST-precise capsules, and uses Gemini 2.5 (or NVIDIA NIM, or your own keys) to write, debug, refactor, and ship code. Every code change is gated by a verification command — no "looks good to me" hand-offs.

Website: try-mesh.com · Docs: try-mesh.com/docs · Source: github.com/dreddi-edit/mesh

Install

npm install -g @trymesh/cli

Requires Node.js ≥ 20. Works on macOS, Linux, and Windows (WSL).

After install, run the golden path:

mesh /start

This runs doctor (env + auth check), indexes your workspace, surfaces a brief, and you're ready for /change.

What you get on install

The mesh interactive REPL (and aliases mesh-agent, mesh-daemon)
71+ slash commands (/change, /index, /doctor, /scaffold, /usage, …)
100+ workspace tools the agent can call (read, write, grep, run-command, AST refactors, timeline, …)
A bench suite + static-bench guard + median runner
A persistent capsule store + task memory under .mesh/
Zero-config Gemini access via the shared Mesh gateway — no API key needed to start

Quick examples

# Interactive session in your project
mesh

# One-shot, sticky session
mesh "explain how the cost router decides between models"

# Force a model
mesh --model pro "refactor invoice.ts to use async/await"

# Inspect tier + capsule + system prompt without running an LLM call
mesh scaffold inspect "rename Status to JobState across all files"

# Run the static compression bench
node $(npm root -g)/@trymesh/cli/bench/static-bench.mjs

Adaptive scaffolding (what makes Mesh different)

Every turn, Mesh classifies the request into a complexity tier:

Tier	When it fires	What's active
trivial	≤ 2 files & ≤ 300 LOC & QA/single-edit intent	Minimal system prompt, no capsule, no workflow state, 4-step budget
small	≤ 10 files, single-intent	Compact path+signatures capsule, lightweight evidence ledger, 8-step budget
complex	Refactor/ship/risk-high/≥ 8 files	Full capsule + reverse-index, workflow state, edit-session tracking, 18-step budget, optional ensemble sampling

You can inspect a decision without an LLM call:

mesh scaffold inspect "explain the auth flow" --capsule --system-prompt

Force a tier via MESH_FORCE_TIER=complex|small|trivial for power-user sessions or debugging.

Configuration

Mesh works out of the box. For your own keys / overrides:

# Use your own Google API key (bypasses shared gateway)
export GOOGLE_API_KEY=your_key_here

# Override default model
export MESH_MODEL_ID=google/gemini-2.5-pro

# Force scaffolding tier
export MESH_FORCE_TIER=complex

# Ensemble sampling for variance-sensitive sessions (1-3)
export MESH_ENSEMBLE_SIZE=2

# Cap session cost
export MESH_MAX_COST_USD=1.00

# Hide per-turn cost footer
export MESH_HIDE_COST=1

Models

Default routing goes to google/gemini-2.5-flash via the Mesh shared gateway. Use --model <alias> to override:

flash / default — Gemini 2.5 Flash (default)
pro — Gemini 2.5 Pro (reasoning-heavy tasks)
lite — Gemini 2.5 Flash-Lite (max throughput)
3-flash, 3-pro — Gemini 3.x previews
qwen3-coder, kimi, mistral-large, deepseek, llama4, nemotron — NVIDIA NIM

Verify your install

mesh doctor

Checks gateway connectivity, auth, native bridge availability, workspace permissions, and capsule store integrity.

License

Built by try-mesh.com.