Product definition

What is Lantide Data?

Lantide Data is a local-first desktop workspace that lets people and AI agents analyze CSV, Excel, Parquet, and databases while keeping the Plan, SQL, validation evidence, Report, and activity together outside any single chat session.

Codex, Claude Code, and other MCP-compatible agents can use Lantide as the local analysis and review surface. The external agent handles reasoning and orchestration. Lantide keeps execution boundaries and reusable analysis artifacts visible.

Download Lantide Data See use cases

The problem it solves

AI can produce an answer quickly, but a later session may not retain the exact files, metric definitions, joins, checks, and limitations that produced it. Lantide keeps those parts in a workspace that can be reviewed and reused.

What stays together

Plan

The question, grain, metric definitions, joins, and validation requirements.

SQL and evidence

The executable query logic, intermediate checks, and source context behind the result.

Report

Findings, limitations, and the explanation a reviewer can inspect or hand off.

Activity and lineage

Agent actions, approvals, and the relationship between current evidence and stopped or replaced Plans.

Ways of working

The difference is not whether an Agent can produce an answer

The difference is how the work can be reviewed, retained, and continued after the answer is produced.

Dimension	Traditional analysis workflow	General Agent without a dedicated analysis workflow	Agent + Lantide
Primary unit of work	SQL, notebooks, reports, and communication records	Conversations, task files, and a final answer	A project that keeps the Plan, SQL evidence, Report, and Activity together
Review before execution	Depends on team process; review may begin with SQL or results	Depends on prompts, files, and tool permissions	Review and annotate the Plan before a person starts formal Execute
Evidence behind results	A reviewer may need to reconstruct context from several tools	Depends on which files, logs, and tool results the session retains	Executed Plan, query evidence, checks, and Report remain in one workspace
Traceability	Depends on Git, notebooks, BI history, and documentation practices	Context may be split across sessions or project files	Trace from a Report to formal evidence and stopped or replaced Plan history
Handoff and reuse	Depends on documentation quality and the original analyst	Depends on retained context and file structure	People and Agents can reopen project artifacts and Reference Docs across sessions
Data and execution boundary	Depends on the existing analytics stack	Depends on the Agent environment and granted access	Local DuckDB workspace with local files, databases, and MCP data sources
Best fit	Highly customized, human-led analytical work	Fast exploration and low-risk, one-off questions	Analysis likely to be reviewed, questioned, rerun, or handed off

Primary unit of work

Traditional analysis workflow: SQL, notebooks, reports, and communication records
General Agent without a dedicated analysis workflow: Conversations, task files, and a final answer
Agent + Lantide: A project that keeps the Plan, SQL evidence, Report, and Activity together

Review before execution

Traditional analysis workflow: Depends on team process; review may begin with SQL or results
General Agent without a dedicated analysis workflow: Depends on prompts, files, and tool permissions
Agent + Lantide: Review and annotate the Plan before a person starts formal Execute

Evidence behind results

Traditional analysis workflow: A reviewer may need to reconstruct context from several tools
General Agent without a dedicated analysis workflow: Depends on which files, logs, and tool results the session retains
Agent + Lantide: Executed Plan, query evidence, checks, and Report remain in one workspace

Traceability

Traditional analysis workflow: Depends on Git, notebooks, BI history, and documentation practices
General Agent without a dedicated analysis workflow: Context may be split across sessions or project files
Agent + Lantide: Trace from a Report to formal evidence and stopped or replaced Plan history

Handoff and reuse

Traditional analysis workflow: Depends on documentation quality and the original analyst
General Agent without a dedicated analysis workflow: Depends on retained context and file structure
Agent + Lantide: People and Agents can reopen project artifacts and Reference Docs across sessions

Data and execution boundary

Traditional analysis workflow: Depends on the existing analytics stack
General Agent without a dedicated analysis workflow: Depends on the Agent environment and granted access
Agent + Lantide: Local DuckDB workspace with local files, databases, and MCP data sources

Best fit

Traditional analysis workflow: Highly customized, human-led analytical work
General Agent without a dedicated analysis workflow: Fast exploration and low-risk, one-off questions
Agent + Lantide: Analysis likely to be reviewed, questioned, rerun, or handed off

This compares typical working patterns, not every product or team. Lantide creates review and traceability checkpoints; it does not guarantee that AI analysis is correct.

When Lantide is useful

Repeating a weekly CSV or Excel analysis with a new dated export.
Preserving business definitions and analysis context across Codex or Claude Code sessions.
Reviewing AI-generated SQL and validation evidence before using a result.
Working with local files while keeping analysis artifacts on the user's machine.

When another tool may be a better fit

Lantide is not intended to replace Excel for workbook formatting, Power Query automation, or established BI dashboards. Local-first also does not mean model requests are always offline. AI features may send necessary context to the model endpoint configured by the user.

See the reusable weekly file workflow →

Choose the next level of detail

Start with real scenarios, or inspect the workflow and engineering capabilities behind the product.

See use cases Explore all features