AI infrastructure for the agentic economy.
General Liquidity is an applied product and research lab. We build the infrastructure that lets software agents act in financial markets and the wider economy: environments to train them, benchmarks to prove they are good, identity and governance protocols that make them safe to delegate to, and verifiable records of what they did. On top of that infrastructure we ship Gordon, a frontier trading agent already in operators' hands. The two sides reinforce each other. Gordon shows us where the infrastructure has to be stronger, and the infrastructure is what makes an agent trustworthy enough to hand real capital.
Visit our website → · Meet Gordon → · Read the research →
Most of the economy is still hard for software to touch directly. Markets are the exception. Prices update continuously, venue state is observable, and the path from decision to action to outcome is short and difficult to fake. That is why we started in trading. It is the most demanding place to run an agent honestly, and the fastest place to find out whether one actually works.
The same forces are moving outward. As money becomes programmable and settlement shifts onto rails that software can call natively, agents stop being passive tools and begin acting as economic participants: holding budgets, buying services, and settling with one another. That world will need infrastructure it does not have yet. An agent has to prove who it is, carry spending authority it cannot exceed, stay governable when it misbehaves, and leave a record of its decisions that an outside party can check. We build that infrastructure, and keep it open where it is meant to be shared.
| Project | Description |
|---|---|
| Gordon | A frontier trading agent. It scans markets, reasons through setups, writes an explicit plan, shows you the orders before they go out, and executes only inside the limits you set. Terminal-native, model-agnostic, and built capital-safety first, across crypto and equity venues. |
The environments and benchmarks behind the agent.
| Repository | Description | Stars | Issues | PRs |
|---|---|---|---|---|
| OpenOutcry | A point-in-time market environment for training and evaluating trading agents without lookahead leakage. Every scenario is reconstructed from a seed so runs stay reproducible, with native bindings for Rust, Python, and WebAssembly. |  |
 |
 |
| SharpeBench | A benchmark that refuses to reward luck. It scores agents on the Sharpe ratio that survives deflation for the number of strategies tried, and asks them to commit before the evaluation window so a result cannot be fit after the fact. |  |
|
|
Protocols for agents to identify themselves, hold authority, and stay governable.
| Repository | Description | Stars | Issues | PRs |
|---|---|---|---|---|
| Agent Disclosure Protocol | A protocol for an agent to state who it is, what it is permitted to do, and where it came from, in a form another system can check instead of taking on faith. | |
|
|
| AgentWorth | A governance layer for agents that move money: spending mandates they cannot exceed, kill switches, and signed enforcement that leaves an auditable trail. | |
|
|
The verifiable spine both halves are built on.
| Repository | Description | Stars | Issues | PRs |
|---|---|---|---|---|
| Fintrieval | Verifiable, point-in-time memory for the financial agentic economy: a system-of-record for what an agent knew, that it was allowed to act, and that the money reconciled. Bi-temporal recall with no lookahead, cryptographic provenance, and governed writes, over a signed attestation layer anyone can verify offline. | |
|
|
Built from open crates:
sharpebench-sim,sharpebench-stats, andfintrieval-core. In progress: OrderLog, EdgeLint, and TapeTrace.
LinkedIn · contact@generalliquidity.com
Trust is the product. An agent that can move capital is only worth running if you can constrain it, inspect it, and prove what it did, so the safety machinery comes first: permissions that deny by default, an explicit trading constitution, kill switches, and a signed audit trail. We work in the open where the work is meant to be shared, because infrastructure that other people's agents will depend on should be legible and checkable rather than taken on trust. And we describe what exists plainly. Gordon is real and shipping. Much of the rest is research direction, and we would rather say so than imply it is already built.
General Liquidity. Applied product and research lab.