How AI Agents Read the Crypto Market Using PHEME
An AI agent that trades or advises on crypto positions needs more than price data. Price tells you what happened. What an agent actually needs is context — what regime the market is in, how sentiment is shifting, and whether macro conditions are pushing or pulling against the trade thesis. PHEME is built to supply exactly that context, in a form agents can consume directly.
What PHEME emits
Every call to PHEME returns three structured outputs: regime, sentiment, and macro.
Regime classifies the current market condition — whether the market is trending, ranging, risk-on, or under stress. This is not a rule-based label. PHEME ingests live news and runs it through an LLM classification layer, so the regime label reflects what is actually happening in the information environment, not just price action.
Sentiment captures the directional lean of market participants based on the same news ingestion pipeline. This gives agents a signal that leads price rather than follows it.
Macro overlays Polymarket prediction market odds and traditional-market correlations. This means an agent calling PHEME gets a read on whether broader financial conditions are aligned with or diverging from the crypto-specific picture — a distinction that matters when BTC is moving in response to Fed positioning rather than on-chain flows.
Together these three outputs give an agent a structured view of market state that it can reason over, act on, or pass downstream to another agent in a pipeline.
The MCP surface
PHEME is served over the Model Context Protocol at https://mcp.quanturama.com/mcp. MCP is the interface layer that lets AI agents call external tools and data sources in a standardized way. Any agent runtime that supports MCP — including Claude-based agents, LangChain pipelines, and custom agent frameworks — can add PHEME as a callable tool with a single configuration block.
The design decision here matters. Rather than building a custom integration for every agent framework, PHEME sits behind a protocol that the agent ecosystem is already converging on. You configure it once and the three outputs are available as a typed tool call wherever your agent runs.
What this looks like in practice
Consider an agent responsible for sizing positions in a crypto portfolio. Without market context, that agent is working from price signals alone. With PHEME in the tool stack, the same agent can check regime before executing — pulling back in a stressed or ranging regime, leaning in during a confirmed trend. It can weight sentiment as a leading indicator before a move is reflected in price. It can cross-reference the macro overlay to determine whether a crypto-specific signal is likely to hold or get overridden by broader risk-off conditions.
None of this requires the agent to maintain its own news ingestion pipeline, build its own classification model, or construct its own Polymarket integration. PHEME handles the data layer. The agent handles the reasoning and execution.
This separation is what makes PHEME useful as infrastructure rather than a standalone tool. It does one thing — produce a clean, structured market intelligence signal — and exposes it in a way that slots into whatever you are building.
Multi-agent pipelines
PHEME is also composable in multi-agent architectures. A routing agent can call PHEME to assess regime and then decide which specialist agent to invoke — a trend-following agent in a bullish regime, a volatility-management agent in a stressed one. The regime, sentiment, and macro outputs function as shared state that any agent in the pipeline can read from a single source of truth rather than each agent maintaining its own market model.
This reduces inconsistency across agent decisions and keeps the market intelligence layer separate from the decision layer — a cleaner architecture than embedding market logic inside individual agents.
Getting started
If you are building an agent that needs structured market context, the MCP server is the fastest path to integration. The endpoint is live and the three outputs are available on every call.
Add PHEME to your MCP config: https://mcp.quanturama.com/mcp