# OpenAlice: a local workspace for AI trading agents

> OpenAlice wires the agent CLIs you already run into market data, research workspaces and broker accounts. The desktop build is the intended entry point; the trading side is beta and the README says to start on a simulator.

**TraderAlice/OpenAlice** — Your one-person Wall Street. An AI trading agent covering equities, crypto, commodities, forex, and macro — from research through position entry, ongoing management, to exit.

- Repository: https://github.com/TraderAlice/OpenAlice
- Website: https://openalice.ai
- Stars: 7,199 · Forks: 1,139
- Language: TypeScript
- License: AGPL-3.0
- Published: 2026-09-22 · Updated: 2026-09-22 · Language: en
- Canonical page: https://hysenlabs.com/projects/traderalice-openalice

## What OpenAlice solves for a solo trader running agent CLIs

A developer who uses Claude Code or Codex for market research ends up with the same problem every week: the analysis lives in a terminal scrollback, the price data lives in a browser tab, and the broker lives in a third window. Nothing connects them, so a thesis cannot be re-run, and a decision cannot be traced back to the research that produced it. OpenAlice is built around that gap. The README describes it as an orchestrator that brings agents, market tools, research and trading accounts into one local workspace, with the tagline "Your one-person Wall Street." The audience is narrow and specific: someone who already pays for a model credential, already knows how to drive a CLI agent, and wants the agent's output to be a file they own rather than a message in a hosted thread. It is not aimed at a discretionary trader who wants a charting terminal, and it is not aimed at a team that needs shared, audited infrastructure.

## Workspaces as Git repositories, and the agent loop OpenAlice does not replace

The central mechanism is the Workspace. According to the README, a Workspace is a directory and a Git repository holding conversations, files and history together, and Alice stores its own state under `~/.openalice`. That file-based design is the reason the project works with several agents at once: the README states that native agents keep their own model loops and tool behavior, and that Workspace files, skills and dedicated research projects supply the context. OpenAlice is not a model and not an agent runtime, so it does not route your prompts through a proxy or reimplement tool calling. It supplies market data, fundamentals, news and quantitative tools to whichever agent you point at a Workspace, and it persists what that agent produces. The README notes that available data depends on your configured providers, which is the honest way of saying the tool surface is a function of your subscriptions, not of the project. Above the Workspace sit three specialist surfaces: Chat for general research, AutoQuant for quantitative projects, and Auto Prediction for prediction-market research. Tracked links research to assets and topics, Issues turn follow-up work into scheduled runs that use the Workspace's agent and context, and Inbox collects reports and questions with a path back to the Session that produced them.

## Installing OpenAlice and running a first thesis task

The README points to the desktop app as the shortest path. Download the macOS or Windows build from the releases page, and note that the desktop app includes Pi, so you can connect a model credential or a supported login and start researching. The README states explicitly that you do not need a broker account to begin, which makes the research side usable before any trading integration exists.

If you prefer a terminal or a server, the README points to a CLI installer in `docs/cli-installer.md` and a remote quickstart in `docs/remote-quickstart.md`. Building from source is the third route, and the README gives these commands:

```bash
 git clone https://github.com/TraderAlice/OpenAlice.git
 cd OpenAlice
 pnpm install
 pnpm dev
```

After `pnpm dev`, the terminal prints a UI URL; open it. The README adds that source installs need at least one host agent CLI plus its model credentials, so a source build without an agent CLI will start but have nothing to drive.

The first task the README suggests is a research question rather than a trade. In Ask Alice, choose a Workspace and an agent, then submit something like this:

```text
 Build a thesis on NVDA. Compare fundamentals with sector trends and price
 behavior, save the research in this Workspace, and explain what would prove
 the thesis wrong.
```

What you should see is a research file written into the Workspace directory, which you can then read with Git history attached, continue as a follow-up, turn into a recurring Issue, or take into AutoQuant.

## The execution side is beta, and the README says so twice

Broker connectivity runs through Unified Trading Account, which the README describes as the way to inspect holdings, orders and account state. When AI trading is enabled, agents stage proposed operations and commit their rationale, and you review and approve execution through a mechanism the project calls Trading as Git. That approval step is the design's main safety property: the agent proposes, a human commits. The caution block in the README is unusually direct for a project of this kind. It states that trading execution is beta, advises starting with simulator, paper, demo or testnet accounts, calls OpenAlice experimental software, and says it provides no guarantees of correctness, reliability, profitability or loss prevention. Treat that as the operative constraint. A second limitation is structural rather than a bug: because OpenAlice does not own the model loop, output quality tracks the agent and providers you configure, and the README concedes that available data depends on your configured providers. If your data subscriptions are thin, the tools the agent can call are thin too, and no amount of workspace structure fixes that. A third: the README does not document rollback for a committed operation, so the recovery path after a mistaken execution is not described in the project's published documentation.

## How OpenAlice differs from a hosted LLM trading bot

Most projects in this space are hosted services: you connect a broker through an OAuth flow, the vendor's servers hold the strategy and the keys, and the agent runs on infrastructure you cannot inspect. OpenAlice inverts that. The README states that it runs on your machine by default, that Workspaces are directories and Git repositories you can inspect, edit and back up, and that Alice stores its own state under `~/.openalice` and seals broker credentials at rest. The trade-off is real in both directions. You gain inspectability and portability of your research, and you can switch the underlying agent without migrating a vendor's proprietary strategy format. You lose the operational conveniences of a hosted product: there is no vendor on call when a scheduled Issue fails at 06:00, and the README notes that your chosen agents and data providers still use their configured services, so a local-first tool does not make you locally-contained. The comparison to make is against whatever you are running now. If your current setup is a hosted bot with a fixed strategy, OpenAlice is a different category of tool, closer to a research environment with an execution hook than to an automated strategy runner.

## Licence, upgrade cost and what AGPL-3.0 implies here

OpenAlice is licensed under AGPL-3.0, and the repository carries a `THIRD_PARTY_NOTICES.md` alongside the `LICENSE` file. The practical consequence for most readers is limited, because the intended use is running the software locally for yourself. The obligation matters if you modify OpenAlice and let other people interact with it over a network: the AGPL's source-availability requirement reaches network use in a way the plain GPL does not. If you are considering embedding it in a product, or running a modified build as a service for clients, that is the point to get proper advice rather than reading a summary. On upgrades, the release history shows a fast cadence: v0.94.0 shipped on 2026-09-20, preceded by two betas on 2026-09-19 and 2026-09-20. The repository includes `CHANGELOG.md` and `cliff.toml`, and the build scripts include a migration index step (`build:migration-index`), which suggests state migrations are part of the upgrade path. The README does not describe a supported downgrade procedure, so the safe assumption is that upgrades move forward and that a backup of `~/.openalice` and your Workspaces is what protects you. The last push to the repository was on 2026-09-20.

## Conclusion

Adopt OpenAlice if you already run a supported agent CLI and want its output to land in a Git-backed workspace instead of a chat log, and if you are willing to keep execution on a simulator while the beta warning stands. Do not adopt it as an unattended execution engine: the README states that trading execution is beta and that the software gives no guarantee of correctness, reliability, profitability or loss prevention. Before committing real funds, verify which broker integrations your build actually exposes, confirm what Unified Trading Account shows for your account, and read the data and credentials page to see where state under ~/.openalice lives and how it is backed up.

## FAQ

### What is OpenAlice?

It is an AI orchestrator for trading that runs locally and connects supported agent CLIs to market data, research workspaces and trading accounts. The README describes it as covering research through position entry, ongoing management and exit, with equities, crypto, commodities, forex and macro in scope.

### How do I use OpenAlice for the first time?

The README suggests downloading the desktop app for macOS or Windows, connecting a model credential or supported login, and running a research task in Ask Alice after choosing a Workspace and an agent. You do not need a broker account to start, and a source install instead uses git clone, pnpm install and pnpm dev.

### Does OpenAlice need a broker account?

No. The README states you do not need a broker account to begin researching, and broker connectivity through Unified Trading Account is a separate step. If you do connect one, the README advises starting with simulator, paper, demo or testnet accounts because trading execution is beta.

### Which agents can OpenAlice drive?

The README lists Claude Code, Codex, OpenCode, Pi, Oh My Pi and other supported native agents. It notes that native agents keep their own model loops and tool behavior, so OpenAlice supplies context and tools rather than replacing the agent runtime.

### Where does OpenAlice keep my data?

Workspaces are directories and Git repositories you can inspect and back up, and the README states that Alice stores its own state under ~/.openalice with broker credentials sealed at rest. Your chosen agents and data providers still use their configured services.

## Sources

- [License: AGPL-3.0](https://github.com/TraderAlice/OpenAlice/blob/master/LICENSE)
- [Project website](https://openalice.ai)
- [README](https://github.com/TraderAlice/OpenAlice/blob/master/README.md)
- [Releases](https://github.com/TraderAlice/OpenAlice/releases)
- [TraderAlice/OpenAlice on GitHub](https://github.com/TraderAlice/OpenAlice)

---

Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/traderalice-openalice