Skip to main content
The assistant chains its tools together automatically - schema search, SQL, dashboards, sharing, and memory. You don’t need to tell it which tools to use - just describe what you want.
Everything on this page runs inside the Allium App. To build the same kind of analysis into your own agent, see Allium MCP.

​
Natural language to SQL

The assistant finds the right tables, writes SQL, validates it, and executes it - all from a plain English question. Example prompt:
  1. Schema search - searches Allium’s database schemas to find relevant tables and columns
  2. Documentation lookup - checks Allium docs for data model details and best practices
  3. SQL generation - writes an optimized query using the discovered schemas
  4. Validation - runs EXPLAIN on the query to catch syntax errors before execution
  5. Execution - runs the query on Allium’s Snowflake warehouse (you’ll see a confirmation prompt)
  6. Results - presents the data with a clear summary
Query confirmation - the assistant asks for your approval before executing SQL queries. You’ll see the exact query it wants to run and can approve or reject it.
Compute profile - queries run on the compute profile selected for your session. Pick a larger profile for heavy queries, or a less contended one when the default profile’s queue is busy. See Compute profiles.

​
Terminal dashboards

Before building from scratch, the assistant searches Terminal — Allium’s library of public dashboards built by our analysts. Example prompt:
If an analyst has already answered the question, the assistant points you at that dashboard and can reuse its SQL as the starting point for your own analysis. That’s faster than writing a query, and the numbers match what Allium publishes.
Terminal results are precomputed. The assistant tells you when the numbers were last generated, and links you to the dashboard itself for the full context and visualizations.

​
Dashboards and visualizations

Create interactive dashboards from your data with charts, tables, and configurable layouts. Example prompts:
  1. Runs queries to gather the data
  2. Creates a dashboard in Allium Explorer with appropriate chart types
  3. Configures visualizations - line charts for time series, bar charts for comparisons, tables for detailed data
  4. Returns a link to the dashboard in the Allium App
You can iterate on dashboards in the same conversation - add charts, change layouts, or update queries. Charts come in nine types: chart (line, bar, area), value, table, pie, sankey, treemap, scatter, map, and chord. Dashboards add markdown, spacers, and input controls (dropdowns, switches, date ranges, radio groups) so a dashboard can be filtered by whoever’s reading it. Ask for the shape you want — “make that a sankey”, “add a date range filter” — and the assistant picks the right spec. A single query can carry several visuals, each with its own title and type. The assistant attaches them to the saved query, so they stay in sync when the query is updated.

​
Sharing and export

Share your analysis with the world or export it for offline use.
Creates a shareable URL that anyone can view without logging in — app.allium.so/s/dashboard/{share_id} for a dashboard, app.allium.so/s/{share_id} for a query. Shared queries also get an embed URL (app.allium.so/embed/{share_id}) you can drop into your own site.
Sharing is one-way through the assistant. To stop sharing a dashboard, open it in the App and toggle off public access.

​
Share to X / LinkedIn

The assistant generates a pre-filled post with:
  • A summary of key insights (never raw data)
  • A link to the dashboard
  • Relevant hashtags
You click the generated link to review and post - the assistant never posts on your behalf.

​
PDF reports

Creates a downloadable PDF with tables, charts, and text sections. Supports portrait/landscape orientation, custom titles, and subtitles. Download links are valid for 7 days.

​
Prices, balances, and wallet activity

Ask about token prices, wallet holdings, or transaction history and the assistant queries Allium’s price, balance, and transfer tables for you. Example prompts:
The assistant answers these the same way it answers any other question — it finds the right tables and writes SQL. That means full history and cross-chain aggregation are available, at Explorer’s data freshness (~1 hour).
Need sub-second freshness or no-SQL lookups? That’s what Realtime APIs are for — direct REST endpoints for current prices, balances, DeFi positions, and holdings PnL. Agents can call the same endpoints as tools over Allium MCP.
The assistant can search the web for context that isn’t in Allium’s database. Example prompts:
Useful for:
  • Current events and market news
  • Protocol research and governance updates
  • Context about new projects or tokens
For blockchain-specific data (schemas, tables, on-chain data), the assistant automatically uses Allium’s tools instead of web search.

​
Address lookup

Identify wallets and contracts by looking up labels, names, and project associations. Example prompt:
Returns:
  • Name and label (e.g., “Vitalik Buterin”)
  • Category (e.g., “Individual”, “DEX”, “Bridge”)
  • Project association (e.g., “Uniswap”, “Aave”)
  • Chain information

​
Organization memory

Persistent key-value storage shared across all users in your organization. The assistant remembers preferences, key addresses, and context across sessions. Example prompts:
See the dedicated Organization memory page for details.

​
How the assistant stays accurate

Two mechanisms do most of the work behind the scenes. You don’t invoke them, but knowing they exist explains why the assistant behaves the way it does.
Before writing SQL, the assistant loads Allium’s internal SQL optimization rules: partition pruning, when to use pre-aggregated *.metrics.* tables instead of raw aggregation, EVM address lowercasing, per-chain versus crosschain.* tables, and the chain-specific pitfalls that make a query slow or wrong.The same mechanism supplies the dashboard and visual specs, product guidance, and the canonical list of supported chains. It’s why the assistant cites a dated chain list rather than guessing.These skills are also available to your own agent — see Skills.
For work that would otherwise flood the conversation, the assistant delegates to a focused subagent and keeps only the result:
  • SQL expert - iterating on a complex query until it validates and runs
  • Docs expert - deep documentation research across many pages
  • Dashboard generator - assembling a multi-element dashboard
  • PDF report generator - laying out a report
You’ll see this as a step in the assistant’s progress rather than as a separate conversation.

​
Data pipelines (Beam)

Feature-gated - Beam pipelines are available to organizations with the Custom Transforms feature enabled. Contact [email protected] to request access.
Create real-time data pipelines that stream blockchain events through custom transforms to Kafka topics. Example prompts:
  1. Creates a pipeline config with source (chain + entity), transforms (JavaScript or set filters), and sinks (Kafka topics)
  2. Deploys workers - creates Kafka topics, provisions credentials, and spins up workers
  3. Returns connection details - Kafka credentials and code snippets (Python and TypeScript) to consume the stream
  4. Monitors health - check deployment status and worker health anytime
Pipeline management is fully conversational - create, update, deploy, teardown, and check stats all through chat.

​
Live web apps (Vercel)

Feature-gated - Vercel app deployment requires the feature to be enabled for your organization. Contact [email protected] to request access.
Deploy interactive Next.js web apps from a conversation. The assistant generates the code and deploys it to Vercel. Example prompts:
  • Wallet tracker - track holdings, PnL, and transactions
  • Token analytics - analyze token prices and stats
  • Custom dashboard - SQL-powered dashboard with configurable widgets
  • Wallet flows - Sankey diagram of wallet inflows/outflows
  • Price chart - OHLCV candlestick chart with dynamic coloring and 24h stats
  1. Generates app code - Next.js app with Allium API integration
  2. Deploys to Vercel - builds and deploys the app
  3. Returns a claim URL - you claim ownership with a free Vercel account
  4. Iterate - read, edit, and redeploy files through conversation
After claiming, add your ALLIUM_API_KEY in Vercel project settings to connect your app to live data.