Get This Tool
Isnad
Pricing
- Model
- Free
Summary
When a factual claim surfaces from a multi-agent pipeline, you usually cannot tell which agent generated it, which source it came from, or whether any intermediate step corrupted it — and by the time you need to audit it, the chain is gone. Isnad is a Python library built specifically to solve that traceability gap at the claim level, not the log level.
Isnad attaches provenance metadata to individual claims as they move through agent pipelines, borrowing the narrator-grading logic from classical hadith transmission scholarship to score source reliability at each hop. The vendor describes it as claim-level auditing — you get a trustworthiness grade per claim, not a flat event log. It installs via pip and ships with Docker support and Alembic-managed migrations, which means it slots into existing Python stacks without standing up a separate service. The ceiling appears when your pipeline is not Python-based or when you need a hosted dashboard rather than a library you integrate yourself. Teams outside that boundary are building their own wrapper before they can use the core grading logic.
Bottom line: Isnad earns its place in a Python-native multi-agent research pipeline where claim-level audit trails matter; it is the wrong pick if your team needs a drop-in hosted service or works in a non-Python runtime.
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- Claim-level provenance rather than request-level logging, so when an auditor asks which source a specific fact came from, you can answer with a chain — not a timestamp.
- Narrator reliability grading at each pipeline hop, which means you catch a systematically unreliable agent before its output reaches a downstream model or a human reviewer.
- pip-installable with Alembic-managed persistence, so integration into an existing Python pipeline does not require standing up a separate service or rewriting data access logic.
- Apache-2.0 license with self-host support, which means no vendor lock-in on the provenance store and no usage-based fees as claim volume grows.
- Prometheus integration included in the repository, so provenance metrics can feed into an existing monitoring stack without a separate instrumentation pass.
Cons
Sign in to edit- No non-Python SDK exists, so any pipeline component written in Node, Go, or another runtime cannot call Isnad natively — teams in polyglot stacks end up building an HTTP shim around the library, which is a second system to maintain.
- No hosted dashboard or UI is described in the source material, which means non-engineering stakeholders who need to review claim trust scores must either query the database directly or wait for a team member to build a reporting layer — at which point the integration cost rivals adopting a more opinionated provenance platform.
- The project has four stars and no open pull requests, which signals limited community validation at scale; teams building production systems with strict SLA requirements on the provenance layer will likely migrate to a more established audit framework once volume or compliance stakes rise, because there is no commercial support path and no documented production deployments in the source material.
Community Reviews
Sign in to write a reviewNo reviews yet. Be the first to share your experience.
About
- Platforms
- Python
- API Available
- Yes
- Self-Hosted
- Yes
- Last Updated
- 2026-07-11T04:17:31.781Z
Best For
Who it's for
- Multi-agent knowledge systems
- Research on AI provenance and trust
- Python-based AI pipelines requiring claim-level auditing
What it does well
- Tracking provenance of facts through multi-agent pipelines
- Grading narrator reliability in AI knowledge workflows
- Evaluating claim trustworthiness before serving or reviewing
- Integrating classical epistemology methods into LLM systems
Integrations
Discussion Community
Sign in to commentNo discussion yet. Sign in to start the conversation.
Compare Isnad
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 Isnad free?
- Yes — Isnad is fully free to use. There is no paid tier.
- Is Isnad open source?
- Yes. Isnad is open source.
- Does Isnad have an API?
- Yes. Isnad exposes a developer API. See the official documentation at https://github.com/alizahidraja/isnad for details.
- Can I self-host Isnad?
- Yes. Isnad supports self-hosting on your own infrastructure.
- What platforms does Isnad support?
- Isnad is available on: Python.
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
Isnad is an open-source Python library that tracks the provenance of individual factual claims as they pass through multi-agent knowledge pipelines. The core workflow: each claim carries a structured record of which agent handled it, which source it originated from, and a reliability grade assigned to that narrator — the library’s term for any agent or source in the chain. You integrate it via pip into your existing pipeline code, and it persists provenance records through an Alembic-managed database, making the full transmission chain queryable after the fact.
The differentiating feature is the epistemological grading model. Rather than logging events, Isnad applies a framework adapted from rijāl criticism — the classical Islamic science of evaluating hadith transmitters — to score narrator reliability at each step. This means you are not just recording who touched a claim; you are accumulating a trust signal about whether each handler in the chain has historically produced accurate output. The vendor frames this as grounding AI provenance work in centuries of formalized epistemological method, not just database logging.
Isnad fits narrowly: Python-based pipelines, research contexts where claim-level auditing is a first-class requirement, and teams comfortable integrating a library rather than deploying a managed service. The project is early-stage by star count and contributor activity on GitHub. There is no hosted API, no visual dashboard described in the source material, and no non-Python SDK. A team running agents in a different runtime, or one that needs a provenance dashboard their non-engineering stakeholders can open in a browser, will find the library boundary the wrong shape for their problem.
The repository ships with a Dockerfile, docker-compose configuration, and Prometheus integration, which the vendor includes for observability. A .zenodo.json and CITATION.cff are present, signaling the project targets academic and research audiences alongside production pipelines. Examples and experiments directories exist in the repository structure, though the depth of documentation beyond the README is not fully described in available source material.