Don't choose one AI. Use them all.
Ask once. Compare GPT, Claude, Gemini & Grok — all in one place.
One workspace · every frontier model
Four answers. One prompt. Live.
Stop guessing which model is best. Stream GPT, Claude, Gemini and Grok at the same time over SSE — vote, copy, branch, or export. Everything counts toward one plan allowance, shown as a simple percentage — and chat stays unlimited.
• Leader-based replicated log; one node accepts writes per term, followers mirror entries.
• Elections use randomized timeouts to avoid split votes; majority quorum commits entries.
• Snapshotting compacts the log so new followers catch up in O(state) not O(history).
Gotcha: a stale leader can briefly accept writes after a network partition — always check `currentTerm` before commit.
• A strong leader serializes all writes; the log is the consensus state — replicate first, apply after majority ack.
• Three sub-problems isolated: leader election, log replication, safety (commit only entries from the leader's term).
• Membership changes use joint consensus so cluster reconfiguration stays linearizable.
Gotcha: Reading from the leader still requires a read-index or lease — otherwise you can serve stale data after a partition heal.
• A single elected leader per term replicates an append-only log to followers; entries commit on majority.
• Quorum intersection (any majority overlaps the next) guarantees no two committed values disagree.
• PreVote + ReadIndex extensions are nearly always required in production to avoid disruptive elections.
Gotcha: the original paper hand-waves on disk fsync — losing the log on a single node violates the safety proof.
• Decompose Paxos into three pieces: who's the leader, what's in the log, when is it safe to apply.
• Heartbeats double as a lease — followers reject candidates from stale terms, leaders step down on missed quorum.
• Linearizable reads need either an entry-through-the-log or a ReadIndex with a verified heartbeat round.
Gotcha: reconfiguration with single-server changes can violate safety — use joint consensus.
Smart router
Routes simple queries to cheap models, complex ones to the best — based on request size, intent, and your plan.
Usage ledger
Per-user mutex + Redis HINCRBY reserves usage before the stream starts. Partial responses are always saved — and chat stays unlimited after your allowance.
Stream visibility
Sectional step log: reasoning, RAG retrieval, tool calls, citations — all timestamped. Pause, retry one model without truncating the rest.
Describe an app. Ship it.
Forge is the AI-powered cloud IDE inside ChatterMate. Monaco editor, Docker sandbox per project, xterm.js terminal, live preview, GitHub integration, and three AI modes — Ask, Build, Plan. Auto-checkpoints on every change. One-click deploy to Cloudflare Pages or your VPS.
User signs up, verifies their email, completes their profile, then lands on the dashboard.
Once on the dashboard, first-run tips guide the user through core features…
Turn any sentence into a diagram.
Write the way you always do. Hover a line, hit ⚡, and three seconds later that paragraph is a clean, branded flowchart, timeline, or chart — right inside your document. No second tool, no diagramming syntax, no copy-paste round-trip.
Everything you need. Nothing you don't.
Each surface in ChatterMate is built around one idea: keep you in flow. Switch model, mode, or workspace without losing your place.
Canvas & Artifacts
Code editor, HTML preview, Mermaid diagrams, side-by-side version compare. Streaming AI edits with accept / reject, Ctrl+K palette, attach files into the editor.
- Live SSE edit with diff preview
- Revision history · pin · branch
- Starter templates · slash commands
RAG + Knowledge Base
Drop PDFs, DOCX, PPTX, sheets, images, URLs. Sliding-window chunks, hybrid BM25 + pgvector with RRF re-rank. Inline [N] citations in every answer.
- @mention any doc in your prompt
- Private vs. shared per file
- Live ingestion progress bars
Image generation
Image models from OpenAI, Google and xAI — usage-based billing with atomic reserve & refund on failure. Drop generated images into Canvas or chat instantly.
Enterprise usage gate
Redis ledger with per-user mutex, server-side usage estimate and background sync. 429 rate-limits with Retry-After. Users see usage as a simple % of their plan.
Smart routing
Auto-suggests the right model based on complexity, length, and budget. Three-tier catalog cache (memory · Redis · root) with ETag conditional refresh keeps the list fresh and cheap.
- Complexity score per query
- Plan-aware access map
- 1-click "ask all" fallback
Native mobile
Flutter client with OAuth PKCE, SSE streaming, deep-link callback, animated bubbles. WebSocket transport with heartbeat & auto-reconnect.
- QR preview via Expo Go
- App Store publishing guide
- Background agents · offline retry
Workspaces with guardrails.
Invite by email, pick roles, lock down which models the org can call. Per-team usage dashboards, shared conversations, SSO via CleverCreator. Stripe billing on free / pro / team / enterprise tiers.
from chattermate import ChatterMate
cm = ChatterMate(api_key="cm_live_…")
# Let ChatterMate pick the best model for the prompt
results = await cm.chat.stream(
prompt="Outline a SOC-2 readiness plan.",
model="auto",
stream=True,
)
async for event in results:
if event.type == "token":
print(event.model, event.delta)
elif event.type == "balance_exhausted":
breakOne key. Every model. Every artifact.
Issue scoped cm_live_… keys with usage sparklines, expiry, and pause/resume. SHA-256 hashed, Redis-cached 60s, per-key rate-limited. Same SSE protocol the web app uses.
A workflow built for shipping, not demos.
① Describe
Type a sentence. Drop a Figma link, a PRD, or a GitHub URL. Forge spawns an isolated Docker sandbox with Node 22, Python 3.13, and Postgres.
- • Connect any of 15 OAuth providers in one click
- • Bring your repo or fork a community starter
- • Mobile? Flutter scaffold with QR preview
Start free. Scale predictably.
Every plan includes unlimited chat. Your monthly allowance covers premium features — AI Writer, images, voice, connectors & Forge sandbox hours.
For curious humans.
- ✓Fast-tier models from OpenAI, Anthropic & Google
- ✓10 docs · 100MB knowledge base
- ✓1 Forge sandbox · 4h / day
- ·No team workspaces
For builders & researchers.
- ✓All frontier models
- ✓200 docs · 5GB · @mentions
- ✓3 Forge sandboxes · 12h / day
- ✓Dev API · 1 key
For power users shipping daily.
- ✓All frontier models
- ✓Shared KB · 50GB
- ✓Usage dashboard · SCIM-lite
- ✓Unlimited sandbox hours
- ✓SSO via CleverCreator
VPC-isolated. Your keys, your rules.
- ✓Self-hosted or dedicated tenant
- ✓Bring-your-own model keys
- ✓SAML SSO · SCIM · DPA
- ✓Custom retention · data residency
- ✓99.9% SLA · 24×7 support
♾️ Unlimited chat — after your monthly allowance, keep chatting on our fast model at no extra cost (fair use). Chat only: AI Writer, images, voice, connectors and other premium features use your plan allowance.
All prices in USD. Your usage is shown as a percentage of your plan allowance.
Questions, answered.
How is ChatterMate different from running each chatbot separately?+
You get a single usage allowance, a single document library, a single chat history, and a single API key. Switch models mid-conversation, branch from any answer, and never re-paste context.
Do you train on my data?+
No. Your prompts, documents, and code are never used to train models. We pass them directly to the provider you choose, scoped to your account. Team plans get a signed DPA.
Can I bring my own API keys?+
Yes on Team and Enterprise. We'll route calls through your OpenAI, Anthropic, Google, and xAI keys; ChatterMate still tracks usage against your plan but no provider charges hit our card.
What does Forge actually run my code on?+
Each Forge session spawns an isolated Docker container with Node.js 22, Python 3.13, and Postgres 17 + pgvector. Sandboxes are ephemeral; pin a checkpoint to keep state across sessions.
Can I self-host?+
Yes — Enterprise customers can deploy the entire stack (FastAPI · Postgres + pgvector · Redis · Docker sandboxes) into their own VPC via our docker-compose.prod.yml + Helm charts. Air-gapped is supported with bring-your-own model endpoints.
What happens when I use up my monthly allowance?+
Chat keeps going — every plan includes unlimited chat on our fast model at no extra cost (fair use). Premium features like AI Writer, images, voice and connectors wait for your allowance to reset, or you can upgrade or top up. A reply that is mid-stream is saved, so your work is never lost.
Bring every AI
into one workspace.
Free monthly usage plus unlimited chat, no card. Sign in with CleverCreator or try it as a guest.