Get This Tool
TetherDust
Pricing
- Model
- Free
Summary
AI analytics tools fail the moment someone asks 'but is that SQL actually correct for our schema?' — TetherDust exists because most LLM-to-database pipelines have no way to answer that question.
TetherDust runs inside your infrastructure, connecting MCP servers to your codebase and database documentation so agents generate SQL that can be checked against the actual schema — not guessed. The core workflow chains natural language input through containerized agents that produce SQL, d3.js dashboards, and schema-to-code dependency maps, all inside strict read-only query boundaries. Scheduled reports ship by email or download without exposing write access. RBAC and audit logging are included for teams where data access needs a paper trail. The ceiling appears when you need write operations, or when your branching query logic outgrows what the agent layer can express without custom extensions.
Bottom line: Pick TetherDust when your team needs self-hosted, verifiable SQL generation with an audit trail — plan for a custom extension layer when your analytics workflows require write-back or multi-step conditional logic the agent graph cannot express.
Community Performance Report Card
No community ratings yet. Be the first to rate this tool!
Community Benchmarks Community
Sign in to submit a benchmarkNo community benchmarks yet. Be the first to share a real-world data point.
Pros
Sign in to edit- Documentation-grounded SQL generation verifies queries against your actual schema before they run, so hallucinated column names and wrong table joins surface before they reach your database.
- Full self-hosting via Docker Compose with enforced read-only query boundaries, which means you can deploy on air-gapped or private infrastructure without sending query logic or schema details to an external service.
- RBAC and audit logging are included at the platform level, so every AI-generated query access is traceable — without this, teams typically bolt on audit layers after a compliance incident.
- Schema-to-code dependency mapping updates as schemas evolve, so developers can see the downstream code impact of a migration before it ships rather than debugging broken queries after the fact.
- Provider-agnostic multi-agent support through MCP servers, so swapping the underlying LLM is an infrastructure configuration change rather than a code rewrite.
Cons
Sign in to edit- Read-only query boundaries are enforced by design — any workflow requiring write-back operations, data mutations, or ETL pipelines hits a hard architectural wall, and teams with those requirements move to a database-native AI tool or build a parallel pipeline outside TetherDust.
- Dashboard output targets d3.js specifically, which means customizing visualizations beyond what the agent generates requires direct JavaScript work; teams expecting a drag-and-drop editor or chart type flexibility will find the output layer thin and reach for a dedicated BI tool instead.
- The repository has 4 stars and 9 commits at time of curation — community support, third-party integrations, and documented edge-case handling are sparse, so teams hitting undocumented failure modes are writing the answer themselves rather than finding it in a forum.
- Complex multi-step conditional query logic — branching based on what one agent returns before passing to the next — pushes past what the agent graph handles natively; teams building those workflows add a Python orchestration layer, and at that point they are maintaining TetherDust plus a second system they own entirely.
Community Reviews
Sign in to write a reviewNo reviews yet. Be the first to share your experience.
About
- Platforms
- Docker, self-hosted
- API Available
- Yes
- Self-Hosted
- Yes
- Last Updated
- 2026-06-12T23:51:21.516Z
Best For
Who it's for
- Self-hosted AI analytics in private infrastructure
- Teams needing verifiable SQL generation from code and DB docs
- Organizations requiring audit logging and RBAC for AI data access
- Developers integrating multiple LLM agents with databases
What it does well
- Generate SQL queries and verify them against database schemas
- Build and maintain interactive d3.js dashboards from natural language descriptions
- Map and visualize code-to-database dependencies as schemas evolve
- Create scheduled reports delivered by email or download
- Chat with data using documentation-grounded natural language queries
Integrations
Discussion Community
Sign in to commentNo discussion yet. Sign in to start the conversation.
Compare TetherDust
Spotted incorrect or missing data? Join our community of contributors.
Sign Up to ContributeCommunity Notes & Tips Community
Sign in to contributeBe the first to contribute. General notes, observations, gotchas, and tips from people who use this tool day-to-day.
Frequently Asked Questions
- Is TetherDust free?
- Yes — TetherDust is fully free to use. There is no paid tier.
- Is TetherDust open source?
- Yes. TetherDust is open source.
- Does TetherDust have an API?
- Yes. TetherDust exposes a developer API. See the official documentation at https://github.com/mpospirit-apps/tetherdust for details.
- Can I self-host TetherDust?
- Yes. TetherDust supports self-hosting on your own infrastructure.
- What platforms does TetherDust support?
- TetherDust is available on: Docker, self-hosted.
Hours Saved & ROI Stories Community
Sign in to contributeBe the first to contribute. Concrete time/cost savings, with context. e.g. "Cut my code review backlog from 4h to 45m per week."
Curated lists that include this category
Getting an LLM to write SQL is easy. Getting it to write SQL that is demonstrably correct against your live schema, traceable to the documentation, and auditable by your security team — that is the problem TetherDust is built for. The platform connects containerized MCP servers to your repository documentation and database schemas, so when an agent generates a query or builds a dashboard, the output can be grounded and verified against what actually exists in your code and database. The entire stack deploys via Docker Compose inside your own infrastructure, with no data leaving your perimeter.
The differentiating architecture here is the documentation-grounded agent loop. Rather than sending raw schema dumps to an LLM, TetherDust structures repository and database documentation as context that MCP servers expose to agents. This is what enables the schema-to-code dependency mapping — you can see which parts of your codebase touch which tables as schemas evolve, and agents generating SQL are working from the same documented source of truth your developers use.
TetherDust fits teams building internal analytics on private data — regulated industries, orgs with strict data residency requirements, or engineering teams that cannot afford a hallucinated JOIN to reach production. RBAC and audit logging are built in, so access to AI-generated queries follows the same governance paths as direct database access. Where it breaks: the platform enforces read-only query boundaries by design, so any workflow requiring write-back, ETL, or mutation is out of scope. Teams that need complex conditional branching across multiple agent steps report hitting the ceiling of what the agent graph expresses without adding a custom extension layer — at which point they are maintaining two systems.
Deployment is Docker-based with a provided Compose file and entrypoint script. The project is licensed AGPL-3.0 with no paid tier described in the repository. Multiple LLM agents are supported via the MCP server architecture, and the docs describe d3.js as the dashboard rendering target — meaning dashboard customization beyond what the agent generates requires direct JavaScript work.