Data analyst agent: product recipe

For an agent that answers questions about your data: writes SQL or Python, runs it safely, and returns tables and charts. Every service below is from AgentGid's catalog, with the price and free tier from its official page; the first pick in each part ranks best on free tier, official MCP server, real use and price.

agent.md

What it costs

To start

$0 / month

every first pick has a free tier or is open source (limits apply)

If you outgrow the free tiers

about $189 / month

E2B $150, LangSmith $39; billed by usage: OpenRouter

Model, per 1,000 questions

about $1.2

Claude Haiku 5.5; $9 with Gemini 3.8 Flash; AgentGid estimate at ~8,000 input + 800 output tokens each

How it fits together

Load the user's data (CSV, Parquet or a database) into the analytics database or DuckDB.
The model writes SQL or Python for the question.
Run that code in the sandbox, never on your own server, with a time limit.
Return the table or chart with the code used, so people can check it.

Code sandbox (runs the agent's Python)

E2B first pick MCP	Secure cloud sandboxes (microVMs) where agents run code, start in under a second. Free: Hobby: $100 one-time usage credit, sessions up to 1 hour, 20 concurrent sandboxes · paid from $150/mo · $0.000014/s per vCPU ($0.0504 per vCPU-hour) + $0.0000045/s per GiB RAM; default sandbox 2 vCPU / 4 GiB · in 85 public agent repos `npm install e2b` `pip install e2b` agent.md
Daytona MCP	Fast-starting sandboxes (CPU, GPU, Windows) for running AI-generated code, billed per second. Free: $200 in free compute, no credit card · vCPU $0.0504/hour; memory $0.0162 per GiB-hour; storage $0.000108 per GiB-hour after first 5 GiB free · in 28 public agent repos `npm install @daytonaio/sdk` `pip install daytona` agent.md
Vercel Sandbox MCP	Ephemeral Linux microVMs on Vercel for running untrusted or AI-generated code. Free: Hobby: 5 active-CPU hours, 420 GB-hours memory, 5,000 sandbox creations per month · paid from $20/mo · Active CPU $0.128/hour; provisioned memory $0.0212 per GB-hour; creations $0.60 per 1M (iad1) · in 10 public agent repos `npm install @vercel/sandbox` `pip install vercel` agent.md

Also fits: Modal Sandboxes, Riza, CodeSandbox SDK, Judge0 · all code sandboxes

Analytics database

DuckDB first pick	In-process analytical SQL database for querying Parquet, CSV and JSON files locally. Open source · in 191 public agent repos `npm install @duckdb/node-api` `pip install duckdb` agent.md
Redis	In-memory data store for caching, sessions, and real-time analytics. Free: Up to 30 MB · paid from $5/mo · $0.007 per hour (Essentials), $0.014 per hour (Pro) · in 142 public agent repos `npm install redis` agent.md

Also fits: ClickHouse Cloud, SingleStore · all databases and backends

LLM gateway (one key, many models)

OpenRouter first pick MCP	One OpenAI-compatible API for 500+ models across 80+ providers with routing and fallbacks. Free: 25+ free models, 50 requests/day · Provider list price per token (no markup); fee on credit purchases: 5.5% Standard, 8% Business · in 29 public agent repos `npm install @openrouter/sdk` `pip install openrouter` agent.md

Also fits: LiteLLM, Vercel AI Gateway, Cloudflare AI Gateway, Portkey · all llm gateways and routers

Traces and evals

LangSmith first pick MCP	Tracing, evaluation and monitoring of LLM apps and agents, including prompt and dataset management. Free: Developer plan: 1 seat, up to 5k base traces / mo, then pay-as-you-go · paid from $39/mo · Plus: $39 / seat per month incl. up to 10k base traces / mo; trace overage is pay-as-you-go (rate not listed … · in 678 public agent repos `npm install langsmith` `pip install langsmith` agent.md

Also fits: LangWatch, W&B Weave, Ragas, Pydantic Logfire · all llm observability and evals

Model

Default: fast and cheap	Claude Haiku 5.5 · $0.1 input / $0.5 output per 1M tokens · or GPT-6 Luna, MiMo V2.6 Flash
Harder questions	Gemini 3.8 Flash · $0.75 input / $3.75 output per 1M tokens · or Kimi K2.7 Code, DeepSeek V4 Pro
Hardest work	Kimi K3 · $3 input / $15 output per 1M tokens · or Claude Opus 5.5, Claude Fable 5.1

The cheapest model of each price tier that can call tools, from model prices. Try the default first and move up only for the questions it gets wrong.

Or start from a ready platform

Pyod	A Python library for anomaly detection across tabular, time series, graph, text, image, and audio data. 60+ detectors, benchmark-backed ADEngine orchestration,… Free tier
Hex	A data notebook and analytics platform whose Notebook Agent writes and runs SQL and Python to investigate questions, while Threads answers natural-language dat… Free tier · from $36/mo
PandasAI	A Python library that turns natural-language questions into Python and SQL, runs it against dataframes, databases and data lakes, and returns answers and chart… Free tier
Julius AI	An AI workspace that analyses uploaded files and connected data sources in conversation and produces charts, reports, dashboards and presentations. Free tier · from $20/mo

Libraries

DuckDB	Embedded SQL OLAP database for analytical queries and data processing in Python. `pip install duckdb` · MIT · Gid Score 55
pandas	Data manipulation and analysis library with DataFrames for structured data. `pip install pandas` · BSD · Gid Score 87
Polars	High-performance DataFrame library optimized for data manipulation and analysis. `pip install polars` · MIT · Gid Score 68
Plotly	Interactive data visualization library for Python with rich chart and dashboard support. `pip install plotly` · MIT · Gid Score 72
Streamlit	Framework for rapidly building and sharing interactive data apps and dashboards in Python. `pip install streamlit` · Apache-2.0 · Gid Score 76
E2B Code Interpreter (Python)	Stateful code execution environment for AI agents to run and iterate on code safely. `pip install e2b-code-interpreter` · MIT · Gid Score 55

Agent Skills

exploratory-data-analysis K-Dense-AI/scientific-agent-skills	Performs bounded, local exploratory analysis of explicitly supported scientific files. Supports redacted CSV/TSV/JSON profiles; optional NumPy, HDF5, FASTA/FASTQ, and ba… has scripts `npx skills add K-Dense-AI/scientific-agent-skills --skill exploratory-data-analysis`
sql-optimization-patterns wshobson/agents	Master SQL query optimization, indexing strategies, and EXPLAIN analysis to dramatically improve database performance and eliminate slow queries. Use when debugging slow… `npx skills add wshobson/agents --skill sql-optimization-patterns`
jupyter-notebook openai/skills	Use when the user asks to create, scaffold, or edit Jupyter notebooks (`.ipynb`) for experiments, explorations, or tutorials; prefer the bundled templates and run the he… has scripts `npx skills add openai/skills --skill jupyter-notebook`
sql-queries phuryn/pm-skills	Generate SQL queries from natural language descriptions. Supports BigQuery, PostgreSQL, MySQL, and other dialects. Reads database schemas from uploaded diagrams or docum… `npx skills add phuryn/pm-skills --skill sql-queries`
bigquery-migration-sql-translator google/skills	Translates SQL queries, DDL schemas, and stored procedures from legacy database dialects (Teradata, Oracle, Redshift, Snowflake, SparkSQL, HiveQL) to BigQuery GoogleSQL … has scripts `npx skills add google/skills --skill bigquery-migration-sql-translator`
golden-jupyter yusufkaraaslan/Skill_Seekers	Use when testing the golden_jupyter golden build `npx skills add yusufkaraaslan/Skill_Seekers --skill golden-jupyter`

Read a skill before you install it: skills can run scripts.

Coding agent to build it

OpenAI Codex Claude Code OpenCode Pi Compare coding agents

Before you ship

Give the agent read-only database credentials.
Keep API keys on the server, never in the browser or app; set spending limits where the provider offers them.
Read each service's terms for your use: data retention, commercial use, rate limits.
Log model calls (traces) and test with your own real inputs before launch.

The brief

# Product recipe: Data analyst agent (AgentGid)

For an agent that answers questions about your data: writes SQL or Python, runs it safely, and returns tables and charts.

Parts were picked from AgentGid's catalog by rules: free tier, official MCP server, use in public projects built with AI agents, downloads, price. Prices are list prices from the official pages (see each card for the check date); confirm before you commit.

## What it costs
- Start: $0 a month on the free tiers of the first picks (their limits apply).
- If you outgrow the free tiers, the lowest paid plans of the first picks are E2B $150, LangSmith $39 (together about $189/month); usage beyond a plan's credits is billed on top.
- Billed by usage (grows with traffic): OpenRouter.
- Model, default: fast and cheap: about $1.2 per 1,000 questions with Claude Haiku 5.5 (AgentGid estimate at ~8,000 input + 800 output tokens each).
- Model, harder questions: about $9 per 1,000 questions with Gemini 3.8 Flash (AgentGid estimate at ~8,000 input + 800 output tokens each).
- Prices of the picks below were checked on the official pages between 2026-10-09 and 2026-10-10.

## How it fits together
1. Load the user's data (CSV, Parquet or a database) into the analytics database or DuckDB.
2. The model writes SQL or Python for the question.
3. Run that code in the sandbox, never on your own server, with a time limit.
4. Return the table or chart with the code used, so people can check it.

## Code sandbox (runs the agent's Python)
- E2B: free: Hobby: $100 one-time usage credit, sessions up to 1 hour, 20 concurrent sandboxes · paid from $150/mo · usage: $0.000014/s per vCPU ($0.0504 per vCPU-hour) + $0.0000045/s per GiB RAM; default sandbox 2 vCPU / 4 GiB · official MCP @e2b/mcp-server · npm e2b · pip e2b · card https://agentgid.com/apis/e2b/agent.md
- Daytona: free: $200 in free compute, no credit card · usage: vCPU $0.0504/hour; memory $0.0162 per GiB-hour; storage $0.000108 per GiB-hour after first 5 GiB free · official MCP · npm @daytonaio/sdk · pip daytona · card https://agentgid.com/apis/daytona/agent.md
- Vercel Sandbox: free: Hobby: 5 active-CPU hours, 420 GB-hours memory, 5,000 sandbox creations per month · paid from $20/mo · usage: Active CPU $0.128/hour; provisioned memory $0.0212 per GB-hour; creations $0.60 per 1M (iad1) · official MCP https://mcp.vercel.com · npm @vercel/sandbox · pip vercel · card https://agentgid.com/apis/vercel-sandbox/agent.md
- Also fits: Modal Sandboxes, Riza, CodeSandbox SDK, Judge0 (https://agentgid.com/apis/sandbox/)

## Analytics database
- DuckDB: open source · npm @duckdb/node-api · pip duckdb · card https://agentgid.com/apis/duckdb/agent.md
- Redis: free: Up to 30 MB · paid from $5/mo · usage: $0.007 per hour (Essentials), $0.014 per hour (Pro) · npm redis · card https://agentgid.com/apis/redis/agent.md
- Also fits: ClickHouse Cloud, SingleStore (https://agentgid.com/apis/backend-db/)

## LLM gateway (one key, many models)
- OpenRouter: free: 25+ free models, 50 requests/day · usage: Provider list price per token (no markup); fee on credit purchases: 5.5% Standard, 8% Business · official MCP https://mcp.openrouter.ai/mcp · npm @openrouter/sdk · pip openrouter · card https://agentgid.com/apis/openrouter/agent.md
- Also fits: LiteLLM, Vercel AI Gateway, Cloudflare AI Gateway, Portkey (https://agentgid.com/apis/llm-gateway/)

## Traces and evals
- LangSmith: free: Developer plan: 1 seat, up to 5k base traces / mo, then pay-as-you-go · paid from $39/mo · usage: Plus: $39 / seat per month incl. up to 10k base traces / mo; trace overage is pay-as-you-go (rate not listed on pricing page); 1 LSU = $1 (Deployment/Fleet/Sandbox usage) · official MCP https://docs.langchain.com/langsmith/langsmith-mcp-server · npm langsmith · pip langsmith · card https://agentgid.com/apis/langsmith/agent.md
- Also fits: LangWatch, W&B Weave, Ragas, Pydantic Logfire (https://agentgid.com/apis/observability/)

## Model (all of these can call tools; prices per 1M tokens)
- Default: fast and cheap: Claude Haiku 5.5 ($0.1 in / $0.5 out) · https://agentgid.com/models/claude-haiku-5-5/agent.md · or GPT-6 Luna, MiMo V2.6 Flash
- Harder questions: Gemini 3.8 Flash ($0.75 in / $3.75 out) · https://agentgid.com/models/gemini-3-8-flash/agent.md · or Kimi K2.7 Code, DeepSeek V4 Pro
- Hardest work: Kimi K3 ($3 in / $15 out) · https://agentgid.com/models/kimi-k3/agent.md · or Claude Opus 5.5, Claude Fable 5.1

## Or start from a ready platform
- Pyod: free tier · https://agentgid.com/agents/pyod/agent.md
- Hex: free tier, from $36/mo · https://agentgid.com/agents/hex/agent.md
- PandasAI: free tier · https://agentgid.com/agents/pandasai/agent.md
- Julius AI: free tier, from $20/mo · https://agentgid.com/agents/julius-ai/agent.md

## Libraries
- `pip install duckdb`: DuckDB (MIT) · https://agentgid.com/tools/duckdb/agent.md
- `pip install pandas`: pandas (BSD) · https://agentgid.com/tools/pandas/agent.md
- `pip install polars`: Polars (MIT) · https://agentgid.com/tools/polars/agent.md
- `pip install plotly`: Plotly (MIT) · https://agentgid.com/tools/plotly/agent.md
- `pip install streamlit`: Streamlit (Apache-2.0) · https://agentgid.com/tools/streamlit/agent.md
- `pip install e2b-code-interpreter`: E2B Code Interpreter (Python) (MIT) · https://agentgid.com/tools/e2b-code-interpreter-pypi/agent.md

## Agent Skills (read a skill before you install it: skills can run scripts)
- `npx skills add K-Dense-AI/scientific-agent-skills --skill exploratory-data-analysis`
- `npx skills add wshobson/agents --skill sql-optimization-patterns`
- `npx skills add openai/skills --skill jupyter-notebook`
- `npx skills add phuryn/pm-skills --skill sql-queries`
- `npx skills add google/skills --skill bigquery-migration-sql-translator`
- `npx skills add yusufkaraaslan/Skill_Seekers --skill golden-jupyter`

## Coding agent to build it
- OpenAI Codex: free tier, from $8/mo · https://agentgid.com/agents/openai-codex/agent.md
- Claude Code: from $20/mo · https://agentgid.com/agents/claude-code/agent.md
- OpenCode: free tier, from $10/mo · https://agentgid.com/agents/opencode/agent.md

## Before you ship
- Give the agent read-only database credentials.
- Keep API keys on the server, never in the client; set spending limits where the provider offers them.
- Read each service's terms for your use (data retention, commercial use, rate limits).
- Log model calls (traces) and test with your own real inputs before launch.

Recipe page: https://agentgid.com/build/recipes/data-analyst-agent/ · all product recipes: https://agentgid.com/build/recipes/agent.md

Other recipes

AI assistant over your documents (RAG) Voice agent (phone or web) AI chatbot for a website or app AI support agent for your customers Web research and scraping agent Telegram bot with AI Solana trading or alert bot Video from a prompt Design recipes

Picked by rules from AgentGid's catalog, not paid placement. Prices are list prices from the official pages, checked between 2026-10-09 and 2026-10-10; check them before you commit.