Financial Evidence for AI agents

Read-only MCP and API research infrastructure for AI finance and trading agents. Connect to funding, bank-risk, market-liquidity and economic evidence while keeping source dates, coverage gaps and rights attached to the result. Operated by LIQUILENS PRIVATE LIMITED.

Install

Python 3.10 or newer:

python3 -m pip install 'financial-evidence[workspace]==0.1.11'
financial-evidence-trading doctor
financial-evidence-trading context

The workspace extra supplies the MCP/HTTP dependencies needed by the persistent clients. The base package has no required dependencies and includes the terminal and three-tool stdio router. Public hosted research reads need no data API key. Optional subscriptions and client/model costs have separate terms.

Complete your first review

Version 0.1.11 adds a guided first result, a readable source-cited brief and comparison with your next review. No optional dependencies or model are needed:

python3 -m pip install financial-evidence==0.1.11
financial-evidence start
financial-evidence review funding --output reviews

The saved capture retains the original JSON, source dates, gaps and a hash receipt. At your next useful review, add --previous with that capture's result.json to compare evidence. Installation does not start a scheduler. Follow the first-review guide.

Reuse an MCP connection

For the recurring workflow SDK/CLI currently available in repository source, including saved comparisons and explicit outcome reporting, see the agent workflow guide. The published 0.1.11 package predates this addition.

import asyncio
from financial_evidence.trading import ProductClient

async def main():
    async with ProductClient(mode="test", timeout=15) as client:
        first = await client.call("seiche", "data_health")
        again = await client.call("seiche", "data_health")
        print(first["timing"], again["timing"])

asyncio.run(main())

mode="test" labels verification traffic. Repeated reads reuse the bounded session; failed requests are not automatically replayed. Session reuse does not make source data fresher or grant trading authority.

Read compact source-status updates

import asyncio
from financial_evidence.trading.context import ContextClient

async def main():
    async with ContextClient() as client:
        snapshot = await client.snapshot()
        print(snapshot)

asyncio.run(main())

The public source-status routes are:

• Contract
• Snapshot
• SSE events

These report metadata, not live quotes. Source freshness, source rights and stream liveness are separate. See the Python/JavaScript stream guide for expiry, reset and reconnect handling. The JavaScript companion is available from that guide and is not represented as a Python module.

Connect an MCP client

For hosted financial research, add this Streamable HTTP endpoint:

https://agents.seiche.info/products/financial-evidence/mcp

Start with financial_evidence_agent_query and {"dataset":"money_markets","entity":"USD","limit":3}. Download the product MCP configurations for LiquiLens, Seiche, Undertow, Palimpsest and the other explicitly labelled research, metadata and sandbox surfaces. Each endpoint has its own tool schema.

For the local three-tool research router:

uvx --from financial-evidence==0.1.11 financial-evidence mcp

The existing financial-evidence-mcp command remains available. The local router exposes financial_evidence_topics, financial_evidence_route and financial_evidence_fetch; it is separate from the hosted Research Desk tool set.

Diagnostics and source admission

financial-evidence-trading latency --samples 3 performs labelled synthetic reads. source-access inspects the permission catalog, while feed-pilot runs invented recovery fixtures. Neither obtains a feed licence, executes a trade or establishes a production latency guarantee. Preserve unavailable and restricted results.

Setup guide · Source · Issues · Licence