# AnswerOverflow: making Discord thread knowledge findable on the web and in AI agents

> AnswerOverflow adds an indexing layer to Discord servers by converting selected channel threads into web pages that search engines can crawl. It is a TypeScript monorepo using Convex, Next.js 16 and Effect, and the hosted service at answeroverflow.com runs larger communities including Cloudflare and Nuxt.

**AnswerOverflow/AnswerOverflow** — Powering Discord content discovery by making web pages from threads for some of the largest servers, including Valorant, Cloudflare, C#, and Nuxt.

- Repository: https://github.com/AnswerOverflow/AnswerOverflow
- Website: https://www.answeroverflow.com/
- Stars: 2,022 · Forks: 102
- Language: TypeScript
- License: MIT
- Published: 2026-10-09 · Updated: 2026-10-09 · Language: en
- Canonical page: https://hysenlabs.com/projects/answeroverflow-answeroverflow

## Bot joins Discord, channels are selected, and threads become indexed web pages

AnswerOverflow addresses a specific failure mode of Discord as a knowledge base: questions get answered in threads, but those answers are invisible to search engines and inaccessible to anyone who was not in the server at the time. As the README describes, the four-step flow starts when the bot joins a server: an admin selects which channels to index, the platform creates web pages from the threads in those channels, and those pages become indexable by Google and other search engines.

Large communities running this system are named in the project description: Valorant, Cloudflare, C# and Nuxt. Beyond Google, MCP is named alongside ChatGPT as a discovery surface for indexed content. A .mcp.json file at the repository root confirms that Model Context Protocol integration is a first-class feature rather than an external addition, and packages/agent/ holds the AI agent system as its own package within the monorepo.

A `.cursor/` directory and `.vscode/` directory are both present at the repository root, providing editor configuration for at least two tools. The `features/` directory at the top level sits alongside apps/ and packages/ in the monorepo structure, representing a third organizational layer for code that does not fit cleanly into a single application or shared package.

## Convex handles the database, auth and real-time backend as a single managed service

Convex sits at the center of the backend, listed under packages/database alongside auth in the project structure. A backend-as-a-service, Convex combines a document database, server functions, real-time sync and authentication in one platform, which removes the need to manage separate database, cache and auth services.

Better Auth handles authentication at the application level on top of the Convex backend. Next.js 16 handles the web front-end, and the Discord bot uses discord.js. Taken together, this forms a T3-stack-adjacent architecture: TypeScript throughout, with the database layer handled by a managed service.

Swapping Convex for another database is not a documented option. Packages/database owns the Convex schema, mutations and queries, and any migration would require rewriting that package plus adjusting every consumer in the monorepo. For teams evaluating self-hosting, Convex's own pricing and terms apply to any deployment, a topic the repository does not cover.

## Effect + TypeScript models errors at the type level inside the Discord bot

Alongside discord.js, the bot package uses Effect, a functional programming library that makes errors, async operations and external dependencies explicit in function signatures rather than relying on runtime exceptions. A function that might fail in Effect code says so in its return type, and calling it requires handling both paths.

For a team already familiar with Effect, this makes failure modes visible in types rather than discovered through runtime logs. For a team unfamiliar with it, the learning cost is real. Effect uses a pipe-based composition model, and the testing approach differs from standard TypeScript conventions. Bot behavior is what requires modification for custom channel or message handling, and understanding Effect is the prerequisite for that work.

Teams that only want to run AnswerOverflow without modifying bot behavior can treat the bot as a black box and configure it through the settings documented in CONTRIBUTING.md. That lower-cost path limits what is adjustable without reading deeper into the codebase.

## Bun 1.3+, Node 18+ and a Discord bot token are the three prerequisites

Development requires Bun 1.3+, Node 18+ and a Discord bot token. Bot token setup is in CONTRIBUTING.md rather than the README.

```bash
bun install
bun dev
```

`bun dev` maps to `turbo run dev` in the root package.json, using Turborepo to start all workspace packages simultaneously. A cold start brings up the Discord bot from apps/discord-bot/ and the Next.js web application from apps/main-site/ together.

For development with databases running locally, a separate `dev:with:dbs` script exists in package.json: it starts the databases first via `bun scripts/start-dbs.ts`, then runs Turborepo. A docker-compose.yml at the repository root configures Jaeger for distributed tracing, with port 16686 for the Jaeger UI, port 4317 for OTLP gRPC and port 4318 for OTLP HTTP. Jaeger is an optional observability tool for local debugging and not a required dependency for basic development.

## MCP support exposes Discord content to AI agents, with setup details outside the README

Model Context Protocol support exists through .mcp.json at the repository root and the packages/agent/ package. MCP is a protocol that lets AI assistants and coding tools access server-provided resources and tools, and AnswerOverflow uses it to make indexed Discord content available to AI agents. This is distinct from the Google indexing path: MCP provides a structured API layer for programmatic access rather than a crawlable web page.

What the MCP server exposes, which tools it provides and which queries are available are not documented in the README. Documentation at docs.answeroverflow.com and in CONTRIBUTING.md is where setup details live. For teams wanting to connect a coding assistant to their indexed Discord content, .mcp.json is the configuration entry point. An opencode.jsonc file at the repository root and an .opencode/ directory both provide configuration for the OpenCode editor, extending AI assistant support beyond MCP clients.

## Production deployment is not in the README, and Stripe billing adds another setup step

Two-command development setup is well documented. What follows from there to production is not.

A Vercel OSS Program badge and a `vercel env pull .env.production --environment production` script in package.json both point to Vercel as the deployment target for the main site. Discord bot deployment is a separate concern, and it is not covered in the README. Convex backend setup also requires its own account and environment configuration.

Stripe integration is present: a `stripe:listen` script in package.json sets up local webhook forwarding. Stripe billing adds a third service account and configuration step for any self-hosted deployment, and the README does not document those steps. Teams treating this as a self-host-first project should budget time for working through CONTRIBUTING.md and the external documentation before reaching a production state.

## Discourse as a web-first alternative versus AnswerOverflow's indexing-on-top approach

Forum software like Discourse starts web-indexable by design: every question and answer is stored in a database that search engines can crawl from the moment of posting, with no secondary indexing layer required. Moving a community from Discord to Discourse means asking members to adopt a different tool for technical discussion, which carries adoption friction: new accounts, different notification settings and a different thread model all need to become familiar before the community is productive again.

AnswerOverflow takes the opposite position: leave Discord unchanged and add the indexing layer on top, so members continue using the tool they already use for conversation. Ownership is where the trade-off becomes concrete. A self-hosted Discourse instance keeps all pages in a database under the operator's control. Using answeroverflow.com as a service means indexed pages are hosted on their platform, and removing the bot later does not retrieve those pages from that hosting. Self-hosting AnswerOverflow restores that control but adds the Convex, Vercel and bot deployment complexity. Neither path is unconditionally simpler; they make different things easy.

## v1.5 released on 2024-10-28, with Biome 2.1.4 for linting and the last push on 2026-09-20

Version v1.5, released on 2024-10-28 and titled "Clean up codebase & move to bun", is the most recent tagged release. Before that, v1.0 was released on 2023-03-28 and v0.9 on 2023-03-16. A release-please-config.json at the repository root sets up release automation, but no new tag has appeared since 2024-10-28. The last push to the repository was on 2026-09-20, and it is not archived.

Biome at version 2.1.4 (as @biomejs/biome in devDependencies) handles linting and formatting in a single tool, replacing the separate ESLint and Prettier combination common in older TypeScript monorepos. A SECURITY.md file is present, relevant for a project that indexes Discord server content. Licencing is MIT under LICENSE.md. For teams indexing a Discord server they do not administer, Discord's own Terms of Service governs what is permissible, and the repository does not address that question.

## Conclusion

Use AnswerOverflow if your community already lives in Discord and you want its thread content searchable on Google without asking members to change tools. Skip it if you need to control where the indexed pages are hosted, since deploying the full stack requires setting up Convex, a Discord bot and Vercel, with production steps delegated to CONTRIBUTING.md rather than the README. The last tagged release was v1.5 on 2024-10-28, though the last push was on 2026-09-20. Verify Discord's Terms of Service for your use case before indexing a server you do not own.

## FAQ

### What does AnswerOverflow do?

It installs a Discord bot on your server, lets you select which channels to index and converts threads from those channels into web pages that search engines like Google can index. Indexed content is also accessible to AI agents via MCP.

### Is AnswerOverflow self-hosted or a hosted service?

The code is MIT open source and can be self-hosted, but the README does not document production deployment steps; those are in CONTRIBUTING.md and external documentation. A hosted service at answeroverflow.com runs the platform for communities that do not want to manage the infrastructure.

### What database does AnswerOverflow use?

Convex is the database, handling storage, auth and real-time sync as a managed service. Packages/database holds the schema and queries for the Convex backend.

### What are the development requirements for AnswerOverflow?

Bun 1.3+, Node 18+ and a Discord bot token are listed as requirements. Token setup is in CONTRIBUTING.md. Running development starts with `bun install` followed by `bun dev`.

### Does AnswerOverflow support AI agent integration?

Yes, via the Model Context Protocol. A .mcp.json file is present at the repository root and a packages/agent/ package exists in the monorepo. MCP is named alongside Google and ChatGPT as a discovery surface for indexed content.

## Sources

- [AnswerOverflow/AnswerOverflow on GitHub](https://github.com/AnswerOverflow/AnswerOverflow)
- [License: MIT](https://github.com/AnswerOverflow/AnswerOverflow/blob/main/LICENSE)
- [Project website](https://www.answeroverflow.com/)
- [README](https://github.com/AnswerOverflow/AnswerOverflow/blob/main/README.md)
- [Releases](https://github.com/AnswerOverflow/AnswerOverflow/releases)

---

Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/answeroverflow-answeroverflow