GitHub Repos intermediate 3 min read Apr 30, 2026
Public Preview Sign in free for the full digest →

Cossistant: AI Support Agent Inside Your React App

“Your Intercom widget knows nothing about your codebase — Cossistant puts the support agent inside your React app, where your LLM copilots can read and maintain it”

Cossistant: AI Support Agent Inside Your React App
1 Views
0 Likes
0 Bookmarks
Source · github.com

“"I made this out of frustration. I want the support in my SaaS to feel like my product: fully customizable, part of the UI, and able to access my codebase so AI agents can actually do useful things." — Anthony Riera (frenchriera), HN submission https://news.ycombinator.com/item?...”

You know that feeling when your support widget opens a generic third-party chat portal that looks nothing like your product, has no idea what your app actually does, and leaves your AI copilots like Cursor or Claude completely blind to how your support system is configured? You add Intercom or Zendesk, and suddenly your support infrastructure lives entirely outside your codebase — untestable, unversioned, and invisible to every tool in your development workflow. When a customer hits an edge case, your support agent has no programmatic access to their Stripe subscription status, cannot open a Linear ticket automatically, and cannot be maintained by the same LLM that helps you write application code. The support layer becomes a silo the rest of your stack cannot reach.

customer-supportai-agentreactnext-jstypescriptopen-sourcesaas-tooling

You install @cossistant/react (or @cossistant/next for Next.js) and define your support infrastructure as TypeScript code inside your existing repository. The package ships headless hooks and UI primitives that render the support widget as a native part of your application UI. A Hono API backend with tRPC handles requests end-to-end with full type safety, persisting conversation history in Postgres and managing real-time sessions over WebSockets via Redis. The AI agent indexes your documentation and past conversations to auto-build and update FAQs, and it calls pre-built tool integrations — Linear for ticket creation, Stripe for subscription checks, Cal.com for booking — or any custom tool you expose via the API hook. When the agent's confidence is low, it escalates the conversation to a human inbox.

01
In-product widget via npm — you embed the support agent as a React component inside your own UI, not as a third-party overlay your customers have to leave your app to use
02
TypeScript-native config — support logic lives in your repo, so your LLM copilots (Cursor, Claude) can read, understand, and maintain it alongside application code
03
Pre-built integrations with Linear, Stripe, and Cal.com — the agent creates tickets, checks subscriptions, and books calls without any custom glue code
04
Self-learning FAQ generation — the agent crawls your docs and conversation history to auto-build and update FAQs, reducing repeated tier-1 queries over time
05
Human escalation path — the agent hands off to a human inbox when confidence is low, so you do not leave customers stuck with a wrong AI answer
06
Dual package split (@cossistant/react + @cossistant/next) — the core library stays framework-agnostic while Next.js users get first-class utilities without importing framework-specific code into a plain React project
07
AGPL-3.0 free tier — you can run the full stack locally at no cost for non-commercial projects, and evaluate it on the $20/mo Hobby plan before committing to a commercial license
Who it’s for

Developers and small engineering teams (1–10 people) shipping React or Next.js SaaS products on a TypeScript stack who want support infrastructure they can version-control, test, and maintain with the same tooling as their application code. You need Bun runtime, Docker, and an existing React or Next.js app to get started. NOT for: teams on non-React frontends (Vue, Angular, Svelte), companies that need a stable production-grade system with a CHANGELOG and external PR contributions accepted, enterprises with compliance requirements that cannot use AGPL-3.0 without a commercial license, or anyo...

Worth exploring

Worth a 3-day evaluation spike if your stack is React or Next.js with TypeScript and you are actively looking to replace a third-party support widget. The last commit was 2026-04-29 (one day before research date) and the 0.2.0 release dropped 2026-04-20, confirming active development. The free tier and $20/mo Hobby plan make the cost of evaluation near zero. Before shipping to production: get legal sign-off on the AGPL-3.0 commercial license requirement, pin your version in package.json because CONTRIBUTING.md explicitly warns of week-to-week breaking changes, and accept that there is no CHANGELOG.md and external PRs are not accepted. If you need stability today, Chatwoot at 28.9k stars and...

Developer playbook
Tech stack, code snippet, sentiment, alternatives.
PM playbook
Adoption angles, user fit, positioning.
CEO playbook
Traction signals, ROI, build vs buy.
Deep-dive insight
Full long-form analysis, no fluff.
Easy mode
Core idea, fast — when you need the gist.
Pro mode
Technical nuance, edge cases, tradeoffs.
Read the full digest
Go beyond the preview

Deep-dive insight, Easy and Pro modes, plus action playbooks — the full breakdown is one tap away.

Underrated tools. Unfiltered takes.

Read the full digest in the Snaplyze app for deep-dive insight, Easy and Pro modes, and the playbooks you can actually use.

Install Snaplyze →