Resources & Prompts

Beyond callable tools, the Sequence MCP server provides Resources (ambient context data) and Prompts (pre-built analytical workflows) that enhance AI conversations with trading context.

Resources

MCP Resources are data endpoints that clients can read to populate conversation context. Unlike tools that must be explicitly invoked, resources provide background information the AI can reference at any time.

Static Resources

These resources return pre-computed summaries that update on the server side:

These are ideal for Claude Desktop, where the AI reads resources at conversation start to establish context without requiring explicit tool calls.

Dynamic Resource Templates

Parameterized resources that accept a symbol:

To use a template, the AI resolves it with a specific symbol. For example, sequence://ticker/BTC-USD returns the full BTC-USD ticker from the DataEngine.

Prompts

MCP Prompts are pre-defined analytical workflows that structure the AI's approach to common tasks. When invoked, a prompt injects a system message with specific instructions and required tool calls, guiding the AI through a multi-step analysis.

analyze_market

Arguments: symbol (required)

Performs comprehensive market analysis for a trading pair:

1. Fetches current NBBO and spread across all venues
2. Retrieves L2 order book depth and liquidity profile
3. Pulls recent trades for short-term flow analysis
4. Gets DataEngine ticker stats (volatility, depth, dispersion)
5. Checks for arbitrage opportunities
6. Synthesizes findings into a structured market report

Example invocation:

Use the analyze_market prompt for ETH-USDT

The AI returns a report covering current pricing, liquidity depth, spread quality, recent flow direction, volatility regime, and any cross-venue opportunities.

pre_trade_check

Arguments: symbol (required), side (required), size (required)

Evaluates conditions before placing a trade:

1. Checks current NBBO and spread
2. Estimates market impact for the requested size using order book depth
3. Reviews venue-level liquidity to predict likely SOR routing
4. Fetches fee schedules for cost estimation
5. Verifies sufficient balance across relevant venues
6. Provides a go/no-go recommendation with expected costs

Example invocation:

Run a pre-trade check for buying 2 BTC-USD

The AI assesses whether the trade is viable, which venues will likely fill it, estimated slippage, and total expected cost including fees.

recon_report

Arguments: none

Generates a data quality reconciliation report:

1. Fetches the CoinGecko reconciliation data
2. Identifies symbols with price drift
3. Checks venue connectivity for stale feeds
4. Reviews cross-venue dispersion metrics
5. Flags any data integrity concerns
6. Summarizes overall data quality health

Example invocation:

Run the recon report

The AI produces a data quality dashboard showing which symbols are in sync, which have drift, and whether any venue feeds are stale or disconnected.

research_workflow

Arguments: none

Step-by-step guide for running the full research pipeline:

1. Check data coverage (what pairs/venues have data?)
2. Materialize profiles for the latest window — or, in read-only mode, list available snapshots and pick the freshest usable one
3. Scan for high-opportunity pairs using structured JSON filters
4. Deep-profile the top candidates
5. Evaluate signal features for predictive power (read-write) or monitor drift on deployed bundles (read-only)
6. Summarize findings with actionable recommendations

Example invocation:

Run the research workflow

Best invoked at the start of a research session — it primes the AI with the canonical research playbook and the right ordering of scan / profile / feature-eval / monitor calls.

How Resources and Prompts Work Together

In a typical Claude conversation:

1. Resources provide ambient context -- the AI knows your portfolio state and market overview without being asked
2. Prompts structure complex workflows -- instead of asking multiple ad-hoc questions, invoke a prompt for a thorough analysis
3. Tools fill in specific details on demand -- the AI calls individual tools as needed during conversation