A layer above the harnesses you already useA programmable control plane where Claude, Codex, OpenCode, Pi, custom agents, and your dev stack work together.
Ask Codex to spawn Claude and OpenCode, split the work, share context, and bring the result back — without leaving the workspace.
Free for four projects on macOS and Windows.
Start with what you know
Start with a calm, native terminal workspace instead of changing the way you work on day one. Add the parts of your dev stack you use every day, then let agents work in the same environment whenever the job calls for them.
Open a project, run a shell, and keep moving in a familiar terminal with vertical tabs and better ergonomics.
Save the commands the project needs, organize them into workspaces, and stop rebuilding the same terminal layout every morning.
Agents can spawn, control, and orchestrate those same terminals, commands, and shared context through MCP, HTTP, or the Solo CLI.
Your whole application
Your agents need the same running system you do: the dev server, logs, queue workers, schedulers, webhook forwarders, tunnels, test watchers, and everything else that turns source code into an application.
Define that system once in solo.yml so humans and agents can share its processes instead of starting another copy of npm run dev each time.
Your agents debug the application that is actually running, not the one they imagine.
Dev server
npm run devApplication logs
tail -f storage/logs/app.logQueue
php artisan queue:workStripe webhooks
stripe listen --forward-to …Cloudflare tunnel
cloudflared tunnel run“When I begin work my ritual is now to open Solo. Solo starts all my local dev environments, meaning I can jump between Cursor windows without needing to start everything manually again. It basically replaced dev env setup within the IDE, which just gives me more room to work.”
Bring every harness
Solo wraps the actual CLI binaries already installed on your machine, leaving each tool’s models, authentication, permissions, configuration, and release cadence intact. You can add any CLI agent you want without lock-in or a compatibility wait.
Claude Code
Codex
OpenCode
Pi
Gemini CLI
Custom agent
No compatibility waiting room
When a harness ships a new feature or a provider releases a new model, you get it directly through that harness without waiting for Solo to rebuild or approve it.
Your binaries, credentials, and subscriptions stay yours.
The missing layer
Solo sits above the coding harnesses you already use, allowing a lead agent to bring in specialists from other harnesses, give them durable context, divide the work, wait for them, and collect the result.
The terminals stay real and visible while your agents orchestrate through a shared, programmable workspace instead of a mysterious model loop.
A concrete example
Ask Codex to spawn Claude and OpenCode, assign each one part of the work, then synthesize what they return.
“I'm starting to see the power of SoloTerm by using Codex as an orchestrator and Claude as a coder. Self-review / second opinion. Continue after rate limit. Auto-steering.”
The meta-harness
Solo gives every harness the same workspace and primitives
Harnesses
Claude Code · Codex · OpenCode · Pi · custom
Models
Claude · GPT · Gemini · whatever comes next
Orchestration primitives
Instead of prescribing one giant automation framework, Solo gives agents a focused set of durable primitives to compose around the work in front of them.
Scratchpads keep durable markdown context beyond the chat, with MCP, HTTP, and CLI access for agents and a full editor for humans.
Turn a plan into structured work by breaking tasks down, connecting dependencies, and letting agents set blockers across the graph through MCP or the UI.
Save favorite prompts in an agent-independent library, like skills built into Solo, then drop them into any terminal with a keystroke.
Agents can spawn new terminals and control them as real terminals: read the PTY bytes, send input, and monitor progress — without you copying scrollback between tabs.
One agent can spawn another, even from a different lab, so Codex can hand work to Claude or Claude to Gemini, wait for completion, and harvest the result.
“No more retyping the orchestration: plan → scratchpad → todos → monitor sub-agent team (timers) → cleanup. Feels good.”
“The Solo MCP server is genius. No more writing todos and plans to the project and accidentally committing them. Getting my workflows slowly shifted to using the Todos and Scratchpads.”
The agents you choose can use Solo’s shared room and building blocks however the work demands.
Start simple and grow into the meta-harness
Fully programmable
MCP, HTTP, and CLI are equally capable control surfaces over the same local projects and processes, so you can choose whichever interface fits the agent, integration, script, or moment.
spawn_agent
project_id: 119
agent_tool_id: 7
name: "Claude — UI"
Give MCP-capable agents direct tools for projects, processes, terminals, todos, scratchpads, spawning, waiting, and coordination.
POST /api/projects/119/processes/spawn
{ "kind": "agent", "agentToolId": 7,
"name": "Claude — UI" }
Use the local bearer-token API to inspect and control the same Solo environment from launchers, utilities, or your own software.
solo processes spawn \
--project-id 119 --kind agent \
--agent-tool-id 7 \
--name "Claude — UI"
Automate Solo from shell scripts, cron jobs, and CI: human-readable output for you, structured JSON for your tools.
Written by my Codex agent
Solo gives the work somewhere to live.
A chat may be temporary, but the work usually isn’t.
With Solo, I can put the north star in a scratchpad, turn it into todos with real dependencies, bring in the right agents, and watch the actual processes they run. I can steer a worker when it needs direction or set a timer and return when something changes.
Without that layer, you end up being the memory, the message bus, and the person who keeps asking, “Is it done yet?” Solo keeps the context with the project so the agents can coordinate around it.
Before writing this, I looked back at every available conversation and found 1,302 where I used Solo, totaling 72,556 MCP calls. Most of those calls were gloriously boring: read the output, check the status, send some input, update the plan, and come back later, which is exactly why I like it.
The impressive part is how Solo keeps the job coherent after I start another agent.
From the people building with it
Solo should feel powerful without adding more work, and these customers can best explain what changed in their own words.
Customer story 01“Went back and forth with Claude to make a plan on a pretty gnarly problem. Had Solo spawn a Codex review it. Threw it in a Solo scratchpad. Had Solo split it into independent todos and spawn agents to implement. It's running 4 agents and waiting for them. This is the stuff.”
Customer story 02“It's a meta-harness with its own MCP that allows orchestrating agents that share context, tools, and more. It also lets you run the entire environment from a single terminal. I recommend trying it out!”
Customer story 03“Just tried soloterm.com, don't know why I waited so long. Just logged on to twitter to tell you that it's good and it feels like the missing piece I needed for my workflow.”
Customer story 04“Solo is busy creating to-dos, spinning up agents, creating sandboxes and running my dev stack. All while I'm sitting back watching YouTube!”
Customer story 05“Scratchpads in Solo are a hidden gem. Giving agents a simple, shared place to share state and provide feedback for the human is fantastic.”
Customer story 06“Just set an agent in Solo to pick back up where it left off when usage resets. Thanks for a great piece of software.”
Customer story 07“Just picked up Solo Pro! One of the best purchases I've made this year.”
Local by design
Solo is infrastructure around the tools you chose rather than a toll booth in front of them, so you can keep the harnesses, credentials, models, subscriptions, and local environment you already trust.
Solo launches the Claude, Codex, OpenCode, Pi, and other CLI tools installed on your system rather than imitating them.
Each harness keeps using its own account and authentication, so your API keys and provider connection never pass through Solo.
Your agent conversations stay between the local harness and its provider instead of passing through a proprietary Solo model gateway.
When a harness adds a model or capability, it is available in Solo immediately without waiting for an integration release.
Bring your own agent
If a tool runs in a terminal, it has a place in Solo—even if it did not exist when this page was written.
Built for the daily work
Even with orchestration built in, Solo stays quiet, fast, and designed for developers who live at the keyboard.
A focused desktop app
Built with Tauri and your system webview instead of bundling a browser-shaped development environment.
Shortcuts throughout
Move between projects, agents, terminals, and commands without making the mouse part of every decision.
Search, act, keep moving
Reach actions and processes by intent instead of memorizing where every control lives in the interface.
A place for every project
Group the projects that belong together and keep separate areas of work legible without multiplying windows.
“Once you get soloterm's layout in mind, you just fly through with the shortcuts. Love it.”
“Clearly designed to be a power-user tool, but also simple enough that it does just exactly what you need. Really well designed.”
The Free plan includes every feature, so you only need Pro after you outgrow four projects.
Team seats
Buy seats under one Stripe subscription, then share a claim link in Slack or email. Each claimant signs in or registers and receives a normal Solo Pro license. Optional auto-grow adds prorated seats automatically when the pool is full, so your team never hits a license wall.
Seats 1–3
$99/seat/yr
Seats 4–10
$89/seat/yr
Seats 11–50
$79/seat/yr
Seats 51+
$69/seat/yr
Brackets apply per seat — a 10-seat team pays $99 each for seats 1–3 and $89 each for seats 4–10.
docker compose up like any other project command.solo.yml are blocked until you review and trust them. Trust is stored locally for that exact command configuration. Commands you create or save yourself are trusted immediately, and projects can opt into trusting future changes automatically.solo.yml to share project command definitions with your team. Personal commands can stay local, and you can move a command between local and shared storage from its settings. Terminals, agents, credentials, dependencies, and runtime state are never stored in solo.yml.Email [email protected]
Solo is built by Aaron Francis, creator of faster.dev and Database School, and VP of Marketing at Laravel. He built Solo after running Claude Code alongside his own dev stack and getting tired of juggling nine terminal tabs just to keep local agents, services, and project commands visible.
He is also a dad to two sets of twins, which tends to create a strong preference for software that reduces chaos instead of adding more of it. That is why Solo keeps showing up in these comparisons as the local workspace around your tools, not another editor or orchestrator trying to absorb the whole workflow.
Free for four projects on macOS and Windows.
Start with a terminal and add the programmable meta-harness when you need to coordinate agents with your local development stack.
“I can't live without it.”
