spyc is a two-pane terminal app: a file commander with vi keybindings on one
side, your coding agents on the other — a single session can hold several. The
agents communicate with the file commander via MCP on a secure local socket.
spy was a tool from my VFX days that I loved dearly. This is its rebirth,
joined with agents (c for claude). spyc, a.k.a. "spicy."
Having found myself back spending most of my time in a terminal, working with
different agents, I was missing a few things:
- a file commander built around vi semantics, and a reliable way to render and work with all the markdown I was suddenly reading
- an easy way to send files and paths to agents without fishing them out of Finder
- a consistent way to keep track of which agent needed attention
- tools to easily fish data back out of agent sessions (pasted images, agent responses, etc.)
I know others have fancy tmux setups that do similar but those don't have the
wonderful construction of spy (yes I know there are other great file commanders
too). I'm also aware there are other agent-management tools, mostly aimed at
herding fleets of agents across VMs and desktops and letting them run. spyc is
on the hands-on end of the spectrum. spyc is a lightweight IDE shell for the
agents you're actively working with supported by an integrated diff to monitor
changes and management of adjacent processes e.g. ssh sessions, other working
shells, etc.
Most of the code was written by agents. My hours went into requirements,
review, and testing. My day job is mostly management at this point, and this
project served as a way to learn firsthand what it takes to maintain a complex,
relatively large codebase with these tools. The time saved on architecture and
coding is quickly absorbed by testing and considering vision. This has turned
out to be enormously fun. Though it remains a truism that the hazards are in
the details; the caves are numerous and the dragons breathe fire.
BSD-3-Clause. No telemetry. macOS, Linux and Windows via WSL.
cargo install spyc, or prebuilt binaries on the releases page.
agentdrek•39m ago
spy was a tool from my VFX days that I loved dearly. This is its rebirth, joined with agents (c for claude). spyc, a.k.a. "spicy."
Having found myself back spending most of my time in a terminal, working with different agents, I was missing a few things:
- a file commander built around vi semantics, and a reliable way to render and work with all the markdown I was suddenly reading
- an easy way to send files and paths to agents without fishing them out of Finder
- a consistent way to keep track of which agent needed attention
- tools to easily fish data back out of agent sessions (pasted images, agent responses, etc.)
I know others have fancy tmux setups that do similar but those don't have the wonderful construction of spy (yes I know there are other great file commanders too). I'm also aware there are other agent-management tools, mostly aimed at herding fleets of agents across VMs and desktops and letting them run. spyc is on the hands-on end of the spectrum. spyc is a lightweight IDE shell for the agents you're actively working with supported by an integrated diff to monitor changes and management of adjacent processes e.g. ssh sessions, other working shells, etc.
Most of the code was written by agents. My hours went into requirements, review, and testing. My day job is mostly management at this point, and this project served as a way to learn firsthand what it takes to maintain a complex, relatively large codebase with these tools. The time saved on architecture and coding is quickly absorbed by testing and considering vision. This has turned out to be enormously fun. Though it remains a truism that the hazards are in the details; the caves are numerous and the dragons breathe fire.
BSD-3-Clause. No telemetry. macOS, Linux and Windows via WSL.
cargo install spyc, or prebuilt binaries on the releases page.
https://github.com/Tripstack-Corp/spyc