Cowrie Research

pragma-primary-sources-2026-04-27

5 min read

Pragma primary-source snapshot

Last reviewed: 2026-04-27 UTC

Project status snapshot

What the official materials say

1) Homepage / current framing

  • URL: https://www.pragma.build/
  • Key takeaways:
    • The homepage frames Pragma as “the leading oracle on Starknet,” which is directionally right but narrower than the full current product surface in docs and repos.
    • The strongest current positioning is not just “oracle on Starknet” but a broader verifiable-data stack with multiple delivery surfaces.
    • The live docs and repos show three especially important surfaces: Starknet onchain oracle feeds, offchain API/WebSocket products, and a Miden-native oracle implementation.

2) Docs index / overall product map

  • URL: https://docs.pragma.build/llms.txt
  • Key takeaways:
    • The docs index makes the current scope legible quickly: Starknet docs, API reference docs, WebSocket docs, onchain/offchain API pages, and a dedicated Miden section.
    • This is a strong signal that Pragma should be cataloged as multi-surface data infrastructure rather than as a single smart-contract oracle.
    • The index also shows deprecated Starknet sections for older products like VRF and optimistic oracle, which helps separate current core surfaces from earlier experiments.

3) Starknet introduction

  • URL: https://docs.pragma.build/starknet/introduction.md
  • Key takeaways:
    • Pragma says its data begins with high-quality sources whose signed data is submitted directly to Starknet, reducing reliance on opaque offchain relay infrastructure.
    • The docs emphasize transparent onchain aggregation and flexible downstream feed construction, including spot-style feeds plus computational feeds like TWAP and volatility.
    • This page is the clearest source for understanding Pragma’s native Starknet value proposition.

4) Starknet architecture

  • URL: https://docs.pragma.build/starknet/architecture.md
  • Key takeaways:
    • The architecture docs describe a concrete contract layout built from a publisher registry, oracle proxy, and oracle implementation contracts.
    • Pragma says publishers submit source data, medians are computed per source, and users can choose final aggregation methods when consuming data.
    • The page also exposes deployed contract addresses, audit references, and bug-bounty references, which is helpful for confirming this is live operational infrastructure.

5) Starknet risks page

  • URL: https://docs.pragma.build/starknet/risks.md
  • Key takeaways:
    • Pragma explicitly documents market-manipulation and publisher/source-concentration risks, especially for long-tail assets.
    • The docs classify feeds by low / moderate / high / depreciating risk and recommend back-testing, monitoring, circuit breakers, and secondary data sources.
    • This is a meaningful operational signal: Pragma is not only marketing price feeds, but documenting how integrators should think about production risk.

6) API reference

  • URL: https://docs.pragma.build/api-reference/introduction.md
  • Key takeaways:
    • Pragma documents a separate Node/API product with development and production base URLs, authenticated access, onchain and offchain endpoints, and WebSocket streaming.
    • The docs describe this API as a fast, highly available surface built on Pragma’s network of data providers.
    • This materially broadens the category: Pragma is not only a contract oracle, but also an application-facing market-data API vendor.

7) Miden introduction

  • URL: https://docs.pragma.build/miden/introduction.md
  • Key takeaways:
    • Pragma Miden is described as a native oracle implementation for Miden, designed around Miden’s client-side proving model.
    • Publishers write prices to their own accounts and the oracle computes medians by reading publisher state through Foreign Procedure Invocation.
    • This is especially notable because it shows Pragma adapting its oracle design to a different execution architecture rather than simply porting a Starknet-shaped product everywhere.

8) Pragma v2 blog post

  • URL: https://blog.pragma.build/introducing-pragma-v-2/
  • Key takeaways:
    • The v2 post frames Pragma as permissionless oracle infrastructure meant to reduce triangulation, transfer, and trust costs in data markets.
    • It says developers should be able to build custom feeds in Cairo and deploy data delivery permissionlessly across chains via Hyperlane.
    • Even if parts of this roadmap/product vision may still be evolving, the post is useful for understanding the strategic direction beyond the currently documented v1 Starknet contracts.

9) API launch blog post

  • URL: https://blog.pragma.build/pragma-empowers-starknet-sequencer-with-the-launch-of-the-api/
  • Key takeaways:
    • The blog describes the Pragma API as a low-latency product built on Pragma’s data-provider network, with Starknet as an early user.
    • It explains a concrete first use case: quoting ETH/STRK conversion data for Starknet fee payment flows.
    • This is useful because it ties the API product to a real chain-operations use case rather than only generic data-vendor language.

10) Security blog post

  • URL: https://blog.pragma.build/exploring-pragmas-security/
  • Key takeaways:
    • Pragma highlights an Immunefi bounty, a Zellic audit for Cairo 0 contracts, and a Nethermind audit for Cairo 1 contracts.
    • The post also claims internal testing and monitoring practices, with plans to expose monitoring dashboards for transparency around providers, assets, uptime, and update frequency.
    • This helps explain the project’s emphasis on operational credibility and ecosystem-security posture.

11) GitHub organization and repo surface

Whitepaper / technical-doc status

  • No canonical standalone Pragma whitepaper or litepaper was found in this pass.
  • The strongest current primary-source corpus is the docs site plus the blog and GitHub repos.
  • In practice, the docs index, Starknet architecture/risk pages, API reference, and Miden docs are a better source of truth than any single paper.

Research notes

  • Best current category for the library: oracle / market-data infrastructure / Starknet-native financial-data and API stack.
  • Key categorization insight: Pragma has outgrown the frame of “just a Starknet oracle” and is better understood as a multi-surface data stack with native onchain, API, and Miden-specific implementations.
  • Another useful note for future scouting: the product brand is Pragma, but the code and org history still sit under Astraly Labs, so repo-namespace mismatch is part of the project’s current discoverability story.

Graph View