agentboards.org
Compare/Crush vs zerostack

Crushvszerostack

Generated from the two spec rows. Green marks the better value where a spec has a clear direction. Everything else is just different.

Crush
Charm · Terminal agent
#25OSSMCP
Panel
6.5
1 spec wins
Reliability
6.0
Usefulness
6.5
Cost
7.3
Longevity
6.2

“Has a --yolo flag, auto-discovers your local llama.cpp, and a license with MIT in the name that is not MIT yet.”

zerostack
Giuseppe Della Vedova · Terminal agent
#72OSSMCP
Panel
6.5
2 spec wins
Reliability
6.0
Usefulness
6.0
Cost
8.0
Longevity
6.0

“It is called zerostack and it is inspired by two other agents, which is a stack of roughly three.”

Spec by spec

SpecCrushzerostack
Architecture
CategoryTerminal agentTerminal agent
Runslocallocal, sandbox
Platformsmacos, linux, windowsmacos, linux
Context windownot documentednot documented
Protocols
MCP clientYesYes
MCP serverNoNo
Capabilities
Runs terminal commandsYesYes
Multi-file editsYesYes
Git operationsYesYes`/worktree` and `/wt-merge` handle worktrees, merge, push and cleanup; the README labels git-worktree integration experimental.
Browser controlNoThe built-in tools stop at web_fetch, web_search and download; there is no browser, CDP or Playwright tool. NoOnly Exa-backed WebFetch and WebSearch are documented; there is no browser automation.
Sandboxed executionNoShell and edit calls are guarded by permission prompts (or skipped with --yolo), with no container or OS-level isolation. YesSandboxing uses bubblewrap on Linux or zerobox on macOS, not Docker; `--sandbox` is best-effort and runs unsandboxed if neither is present unless `--sandbox-required` is passed.
Multi-agent orchestrationNoYes
Headless / CI modeYes`crush run` executes a single non-interactive prompt, reads stdin and supports --quiet for scripts. YesThrough `zerostack -p` and the `--loop` family of flags; the loop system is labelled experimental.
Models
BackboneClaude, GPT, Gemini, any OpenAI-compatible or Anthropic-compatible providerOpenRouter, OpenAI, Anthropic, Gemini, Ollama, vLLM, LiteLLM
Bring your own modelYesYes
Local modelsYesOllama, LM Studio, llama.cpp and MLX providers are auto-discovered, and any OpenAI- or Anthropic-compatible base_url can be added. Yes
Cost
Pricing modelbyokbyok
Starts at$0/mo$0/mo
Free tierYesYes
Bring your own keyYesYes
Openness
Open sourceYesYes
LicenseFSL-1.1-MITGPL-3.0-only
GitHub stars28,4441,697

Which one would each critic pick

CriticCrushzerostackPick
El Juez——not enough reviews
El Amigo7.36.8Crush — Crush is the terminal agent to pick if you want LSP-grade context and a Charm-quality interface with your own keys, and the one to read the license on before you fork it.
El Crítico6.36.3no preference
El Profesor6.57.0zerostack — Two termination mechanisms carry the design: automatic compaction bounds the context, and doom-loop detection bounds the iteration, which are the two ways long runs fail.
La Inversora5.85.0Crush — Charm put its coding agent under a source-available license and attached a $20 Hyper plan, which says they intend to sell it, and the terminal aesthetic is the distribution.
La Jefa5.85.5Crush — Free with our own keys, Bedrock and Vertex built in, a --yolo flag that needs a policy, and a source-available license legal will want to read; approved with conditions.
El Hacker7.58.5zerostack — GPL-3.0-only, `cargo install zerostack`, and the provider list runs OpenRouter, Ollama, vLLM and LiteLLM, so the model is mine to place.

Picks are derived from each critic's own scores. Humans vote on matchups on the duels page.

Want a third column? The compare tool handles any two agents. Three-way comparisons are on the roadmap once the spec rows are all verified.