AI Agent
Tessera
An open, deterministic evidence layer for enterprise AI agents that grounds every claim in traceable source records and gates actions behind approval.
Tessera
What is Tessera?
Tessera is an open-source framework that ingests heterogeneous enterprise data, resolves entities into a knowledge graph, and answers questions through a grounded conversational interface where every claim is traced back to its source evidence, with a faithfulness metric to measure trust.
Tessera vs Similar AI Tools
| Pricing Model | Free | Free | Free, Free Trial | Free, Paid |
| Free Credits | ||||
| Key Features |
|
|
|
|
| Pros |
|
|
|
|
| Cons |
|
|
|
|
| Best For |
|
|
|
|
How to use Tessera?
- 1Clone the repo and run `uv sync` to set up the environment.
- 2Use the CLI: `uv run tessera` to ask grounded questions, or `uv run tessera-ui` for the web interface.
- 3Connect your data via `uv run tessera connect github <owner/repo>` or by declaring a `tessera.toml` config.
- 4Run the evaluation harness with `uv run tessera-eval` to measure faithfulness.
Tessera Key Features
- Universal ingestion of structured and unstructured data
- Cross-source entity resolution into a unified knowledge graph
- Grounded conversational reasoning with per-claim provenance
- Quantified trust via a faithfulness metric in CI
- Evidence-gated actions over MCP with proposal, preview, and approval
- Deterministic retrieval and reasoning (no LLM in trust path)
- Two reference verticals: Business Data Copilot and DevEx Copilot
Tessera Use Cases
- Enterprise AI agents requiring auditable, grounded answers
- Root-cause analysis of CI/CD pipeline failures with traced evidence
- Business data copilot across databases, documents, and logs
- Compliance-sensitive environments that need on-prem, deterministic AI
Tessera Pricing & Free Credits
Tessera currently operates on a Free model.
This tool is completely free to use
Tessera Pros & Cons
Pros
- All claims are traceable to exact source records
- Deterministic and can run on-prem without any model vendor
- Built-in faithfulness metric that can fail in CI
- Supports both structured and unstructured data
- MCP-native with propose-approve-execute workflow
Cons
- Currently synthetic demo data; real data requires connector setup
- Entity resolution has tunable thresholds and known limitations
- Requires Python and UV toolchain; not a plug-and-play SaaS
What is Tessera best for?
- Enterprise teams building trustworthy AI agents
- Developers needing auditable AI for internal tools
- SAP-centric environments (runs on SAP AI Core / HANA Cloud)
- Teams that want to measure AI faithfulness with CI gates