Plainbase — Open-source AI tools every business should have

01 — The tools

Three tools. One philosophy.

Each Plainbase tool does one job. Drop it in, host it yourself, pay only for what you use.

ASK

Plainbase Ask Available

Support chatbot widget that points at your docs and answers visitors automatically.

Drop the widget into your site, point it at your knowledge base, and it handles the obvious questions for you. When it doesn't know, or someone asks for a human, it escalates by email.

Read the docs Learn more

Point at any source. Docs sites, Markdown, PDF, internal wikis.
Email escalation built in. Hands off to support when it should, with full conversation context.
Self-hosted. Conversations and embeddings stay on your servers.
One script tag. Embed in five minutes; theme it to match your site.

MENTION

Plainbase Mention In development

Track how AI systems describe — or fail to mention — your brand.

Mention probes ChatGPT, Claude, Perplexity and others with synthetic user queries about your category, surfaces the answers they give, and flags when your brand is misrepresented, missing, or being recommended in the wrong context. The monitoring layer for the post-search online world.

Get notified at launch

Synthetic query generation. Probes models the way a real customer would.
Sentiment & positioning tracking. See whether you're recommended, dismissed, or omitted.
Diffs over time. Detect when answers change as models update.
Competitor coverage. Compare your share-of-voice in LLM answers.

PLAN

Plainbase Plan In development

Project management built around the reality that nobody updates the tool.

Dump work in as text or voice. Plan organises it into tasks, infers dependencies, checks in with the right owners, and re-surfaces priorities when something slips. New projects start with a guided dump-and-assign flow; the AI does the parts humans skip.

Get notified at launch

Frictionless intake. Voice notes and free-form text become structured tasks.
Automatic check-ins. The AI nudges owners; you don't run status meetings.
Live dependency maps. When something slips, downstream work re-prioritises itself.
Guided project setup. Dump everything once; the AI splits, sequences, and assigns.

02 — Our beliefs

Plain by design. Not by accident.

A small set of constraints we hold to across every tool we ship.

01

One tool, one job

No suites trying to be everything. Each tool earns its place by doing one thing well — and nothing else.

02

Open source by default

APGLv3-licensed source. Inspect it, fork it, audit the prompts, swap the model. Your data, your call.

03

Self-hostable, always

Run on your own infrastructure. We don't see your queries, your customers, or your knowledge base.

04

AI at your service

AI-first, but never AI-only. Humans always in control. Observability built in.

03 — Self-host

Up and running in one command.

Every Plainbase tool ships as a single Docker image with a sensible default config. Bring your own model key — Mistral, OpenAI, Anthropic, or anything OpenAI-compatible including local Llama or Mistral.

Single container, single port, no external services required.
Drop-in support for Mistral, OpenAI, Anthropic, Azure, and OpenAI-compatible local models.
SQLite for state, S3-compatible storage for files. That's the dependency list.

~/plainbase-ask

$docker run -p 3000:3000 \

-e -e AI_PROVIDER=mistral \

-e -e AI_API_KEY=$YOUR_KEY \

-e -e ADMIN_PASSWORD=changeme \

ghcr.io/plainbase-dev/plainbase-ask:latest

✓ Plainbase Ask running on :3000