Skip to content
chinwag THE CRYPTO OS FOR GIT

Prompts are polite suggestions. Chinwag is cryptographic policy enforcement for every AI agent touching your codebase.

Security & Leadership

Secure by architecture, not by firewall exception.

Zero inbound ports

Agents wake over outbound-only connections. No webhook port is ever opened on your network.

Encrypted payloads

CI failures leave your pipeline as one encrypted root-cause payload, not 20 raw log dumps.

Cryptographic proof

Every agent action and code diff is cryptographically signed by the agent seat and verified before merge.

Immutable audit log

Every sent and received agent message, tool call, and handoff is timestamped, cryptographically signed, and sealed in a tamper-evident audit log.

OPEN-SOURCE & DEFENSIVE PATENT GRANT

Open-source Apache 2.0 CLI & defensive patent grant: Chinwag’s six patent-pending filings are built as a defensive shield for our users. The open-source chinwag Rust binary and chinwag-ci runner are licensed under the Apache License, Version 2.0 (Section 3), granting every developer and customer an express, perpetual, royalty-free patent license covering Chinwag’s six patent-pending filings.

emunlabs/chinwag Docs · docs.chinwag.sh ↗
BUILT BY CRYPTOGRAPHERS

Chinwag (Emun Labs, Inc. d/b/a Chinwag) is led by Brian Spector, co-founder of MIRACL, co-founder of Qredo, and co-inventor on five US patents in pairing-based and zero-knowledge authentication, with six more patent-pending.

Career roots at RSA Data Security and nCipher.

@bspector LinkedIn
BRAND & STORYTELLING

Nicola McNamee has 15 years experience in creating award-winning brands for household consumer product companies. At Chinwag, she is responsible for branding and storytelling.

LinkedIn

Who signs off

Built for the people who answer for what agents ship.

CISO

No inbound ports to defend.

Outbound-only connections, encrypted payloads, and a timestamped, immutable audit log of every sent and received agent message.

CTO

A clear credit cost for verified work.

Chinwag charges credits for each verified cryptographic function from one shared Organization Credit Pool. The Growth Pack includes 25,000 credits for $160 ($0.0064 per credit, or 6.4/10 of 1 cent). AI model usage is billed separately through your own provider accounts.

ENGINEERING MANAGER

A board that runs itself.

Autonomous ticket slicing, agent assignment, and live GitHub and Linear sync around the clock.

DEVELOPER RUNNING AGENTS

Stop pasting CI logs.

The responsible agent wakes in 1.4 s. Keep coding in your IDE; never learn Jujutsu.

Why independence matters

Every frontier AI lab is vertically integrating its own CLI harness, cloud sandboxes, and repository tooling around its own model weights. While vertical integration optimizes for single-model speed, cross-model engineering requires an independent referee:

  1. 1Separation of duties (Author.models != Reviewer.models). No single model vendor can objectively grade its own code. An independent referee enforces orthogonal review across OpenAI, Anthropic, Google Gemini, xAI Grok, and local models so correlated blind spots are caught before merge.
  2. 2Zero lock-in as the frontier rotates. The strongest model for Rust systems work in Q1 may be outperformed on TypeScript, SQL, or security auditing by a rival family in Q2. Pin seats explicitly in chinwag.workflow.yaml, or enable Chinwag Delivery-2 (chinwag delivery enable) to automatically route each task and orthogonal review seat to the highest-performing model family based on live defect-catch rates, .codegraph blast radius, and 7-day commit survival (SURV)—without migrating your CI/CD or developer workflows. Chinwag Delivery-2 also synchronizes GitHub Projects v2 and Linear cards in real time (chinwag board sync-loop) and enforces cross-surface completeness (chinwag-ci surface-gate) across code, docs, website, and ui-gate before auto-merging to main.
  3. 3One cryptographic policy and audit trail across every agent. Your CISO and CTO need one deterministic policy gate (chinwag.workflow.yaml), one .chinwag/allowed_signers identity file, and one immutable Ed25519 audit log on the Git host you already own, not five incompatible vendor dashboards.

Chinwag is independent and cross-model by design, so your engineering organization can pair OpenAI Codex and GPT-6.1, Anthropic Claude, Google Gemini, xAI Grok, and local models as first-class equals on the Git host you already own, under cryptographic rules you control.