Moltis
Moltis is a self-hosted AI agent server that runs from a single Rust binary on hardware you own.
If your data has to stay on your own hardware, Moltis is one of the few self-hosted agent servers that covers the whole loop: natural-language cron scheduling, sandboxed execution in Docker/Podman/Apple Container/WASM, SSRF-protected browsing, persistent hybrid vector + full-text memory, voice, and eight chat frontends out of one binary. It also extends itself at runtime with skills, hooks, and MCP servers over OAuth 2.1, and it can import an OpenClaw workspace read-only in one click. The catch is that you are the SRE — TOML config, an always-on host, and vault unlock are yours to run. Pick it for control; skip it if you want someone else to page on call.
Verified 3d ago · liveness 74/100 · cite: rightaichoice.com/tools/moltis
- Developers who want a self-hosted AI agent server keeping keys, memory, and files on their own hardware
- Privacy-focused operators replacing a cloud assistant with a locally run, sandboxed agent
- Power users automating one agent across Telegram, WhatsApp, Discord, Slack, Matrix, Nostr, and Teams
- Teams needing webhook-driven automation for GitHub PR review, Stripe payments, and issue triage
- Anyone who wants a fully managed, zero-maintenance cloud assistant with vendor support
- Non-technical users unwilling to touch a terminal, TOML config, or an always-on host
- Buyers who need a large third-party plugin marketplace rather than built-in features
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip Moltis if you have no always-on host to spare, do not want to own TOML config, Docker/Podman, and vault unlock, and would rather pay a vendor to run your agent than run it yourself.
Moltis itself is free, but every cloud provider you connect bills you directly — a long agent run against Anthropic, OpenAI, or Google Gemini shows up on your provider invoice, not on a Moltis one.
Moltis is open source and free to self-host; the real budget line is the always-on host plus your own provider API usage, so cost scales with how much you run rather than with seats. Compared with managed assistants that bundle hosting and model access into a subscription, Moltis shifts spend toward infrastructure and inference you control. It fits developers and small teams who already pay for a box and are comfortable with variable usage bills.
In short
Moltis — Moltis is a self-hosted AI agent server that runs from a single Rust binary on hardware you own. Best for Developers who want a self-hosted AI agent server keeping keys, memory, and files on their own hardware, Privacy-focused operators replacing a cloud assistant with a locally run, sandboxed agent, Power users automating one agent across Telegram, WhatsApp, Discord, Slack, Matrix, Nostr, and Teams. Free to use.
What's new in Moltis
Checked yesterdayAcross the latest 5 updates: 2 feature updates and 3 changelog entries.
Moltis adds max reasoning effort level and configurable default
Adds a max reasoning effort level (#1253) and persists a configurable default reasoning effort (#1266). Also fixes ACP auto-bind backoff and missing-shell reporting.
Moltis drops unavailable Fireworks model, fixes auth and MCP doctor
Removes an unavailable Fireworks model, counts Docker loopback-only deployments as local (#1249), and fixes doctor validation for streamable HTTP MCP servers (#1251).
Moltis replaces shell-based node-host remote execution
Security fix: node-host remote execution now uses structured, allowlisted process invocation instead of shell. Reported by @yubrew.
Moltis patches sandbox container session key sharing
Security fix: session keys no longer share sandbox containers and persisted homes. Reported by @DavidCarliez.
Moltis adds Obscura stealth mode default and Browserless v2 support
Enables Obscura stealth mode by default (#1227), supports Browserless v2 containers (#1229), and delivers scheduled cron output to the originating chat (#1226).
What people actually say about Moltis — is it worth it?
We ran a structured research pass across product reviews, community discussions, and post-purchase forum threads to surface the patterns vendors won't publish themselves. Below: the recurring strengths, the hidden costs people mention most, and the cohort that consistently regrets adopting this tool.
72 mentions across 5 sources (Hacker News, YouTube, Bluesky, GitHub, Lemmy) · researched Jul 14, 2026.
Average across the 5 sources that answered — each source counts once, not each post.
- +Polished, not full of half-baked features like competitors.
- +One-command installation: three steps from zero to chat.
- +Sandboxed execution with Docker or Apple Containers for security.
- +Persistent memory with vector and full-text search.
- +Multi-provider LLM support: 20+ cloud and local models.
- −Sandbox restrictions cannot be disabled, limiting some use cases.
- −Self-modifying skills raise security concerns among community.
- −Voice and scheduling features lack extensive real-world feedback.
- −Small community means slower support and fewer integrations.
- −No mobile app or managed cloud tier available.
- • Self-hosting costs: hardware, electricity, bandwidth for local LLMs if not using cloud APIs.
Viability Score
How well maintained and how widely used is Moltis? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: October 2026
How we score →Key Features
- Single self-contained Rust binary with no runtime dependencies
- Install via curl, Homebrew, Docker, .deb/.rpm/Arch/Snap/AppImage, or cargo
- Persistent memory with hybrid vector plus full-text search using local GGUF embeddings
- zvec vector database memory backend
- Natural-language cron scheduling that delivers output to the originating chat
- Sandboxed code execution in Docker, Podman, Apple Container, or WASM
- SSRF-protected web browsing that blocks loopback, private, and link-local IPs
- Obscura stealth browsing mode enabled by default
- Browserless v2 container support for remote browser sessions
- Voice input and output via multiple cloud and local TTS/STT providers
- Multi-channel messaging across Web UI, Telegram, WhatsApp, Discord, Slack, Matrix, Nostr, and Teams
- GraphQL API and JSON-RPC interface
- PWA push notifications
- Runtime creation of skills, hooks, and MCP tool servers (stdio or HTTP/SSE, OAuth 2.1)
- Webhook triggers for GitHub PRs, Stripe payments, and arbitrary endpoints
About Moltis
Moltis is a self-hosted AI agent server you install with one curl command and run on a Mac Mini, a Raspberry Pi, or any server you own. It ships as a single self-contained Rust binary with no runtime dependencies, and it is not a chatbot wrapper or a cloud service tied to one provider. Open localhost:13131, paste a provider key — or use zero-config OAuth for GitHub Copilot, OpenAI Codex, or Kimi Code — pick a model, and you have an always-on agent that keeps context across sessions. The core work is automation. Recurring jobs are scheduled in natural language and their output is delivered back to the chat it originated from. Code runs sandboxed inside Docker, Podman, Apple Container, or WASM, web fetches are SSRF-protected, and browser sessions can run in isolated containers with Obscura stealth mode on by default. Webhooks route GitHub PRs, Stripe payments, or any endpoint into the agent, durable calendar and email connectors handle reminders, and MCP tool servers run over stdio or HTTP/SSE with OAuth 2.1 and auto-restart. You reach it wherever you already talk: Web UI, Telegram, WhatsApp, Discord, Slack, Matrix, Nostr, Microsoft Teams, plus GraphQL and JSON-RPC APIs and PWA push notifications. Voice input and output work with cloud and local TTS/STT providers. Memory uses hybrid vector plus full-text search with local GGUF embeddings, and more than 20 providers are supported including Anthropic, OpenAI, Google Gemini, Mistral, DeepSeek, Ollama, LM Studio, local GGUF/MLX models, and any OpenAI-compatible endpoint. Moltis extends itself at runtime with skills, hooks, and MCP tool servers, and it can create multiple agent personas with distinct capabilities through a full CRUD management UI with session branching and hot-reload. Security is the pitch: an encryption-at-rest vault, passkeys/WebAuthn, scoped API keys, human-in-the-loop approval, and an allowlisted process invocation path for remote node hosts. The tradeoff versus a managed assistant is that you own the ops — an always-on host, TOML config, and CLI comfort. Prefer ChatGPT or Claude with zero maintenance? This is the opposite deal.
Behind the Verdict
Moltis occupies a specific and honest niche: it is the agent you run, not the agent you rent. The install story is the strongest part of the pitch — one curl command, or Homebrew, Docker, .deb/.rpm/Arch/Snap/AppImage, or cargo — and what lands on your machine is a single self-contained Rust binary with no runtime dependencies. Open localhost:13131, paste a provider key or use zero-config OAuth for GitHub Copilot, OpenAI Codex, or Kimi Code, and you are running. Where it earns its keep is breadth of workflow in one process. Recurring cron jobs are written in natural language and delivered back to the originating chat — the August 2026 releases explicitly fixed channel-context preservation here, which is the kind of plumbing detail that separates a real scheduler from a demo. Code runs sandboxed in Docker, Podman, Apple Container, or WASM, with a full-filesystem mode available if you want it. Browsing is SSRF-protected against loopback, private, and link-local addresses, and the August 2026 releases turned Obscura stealth mode on by default and added Browserless v2 container support. Webhooks pull GitHub PRs, Stripe payments, or arbitrary endpoints into the agent. CalDAV plus durable calendar and email connectors handle events and reminders. MCP tool servers run over stdio or HTTP/SSE with OAuth 2.1 and auto-restart. Channels are unusually wide for a self-hosted project: Web UI, Telegram, WhatsApp, Discord, Slack, Matrix, Nostr, and Microsoft Teams, plus GraphQL and JSON-RPC APIs and PWA push notifications. Slack saw substantial work in 2026 — per-message acknowledgment reactions, live task cards, Block Kit, reconnect supervision — and Nostr gained NIP-29 group chat. Voice runs through cloud and local TTS/STT providers, and memory uses hybrid vector plus full-text search with local GGUF embeddings, with a zvec vector database backend added in August 2026. Model support now spans GPT-5.6 Luna routing coverage alongside Anthropic, OpenAI, Google Gemini, Mistral, DeepSeek, Fireworks, Cerebras, MiniMax, Moonshot, Z.AI, Venice, and NEAR AI Cloud, with Ollama, LM Studio, local GGUF/MLX, Hugging Face, and any OpenAI-compatible endpoint on the local side. Reasoning effort is configurable, including a max level added in September 2026. Security is the differentiator the README leans on, and the changelog shows the team taking reports seriously: a September 2026 fix stopped session keys from sharing sandbox containers and persisted homes, and another replaced shell-based node-host remote execution with structured, allowlisted process invocation — both credited to external reporters. Vault unlock and recovery now require authentication over HTTP. Skills and hooks are gated behind a per-account operators list for privileged tools. Tool object schemas are kept OpenAI-safe. There is also an encryption-at-rest vault for API keys, passkeys/WebAuthn, scoped API keys, and human-in-the-loop approval. The genuine costs are operational. You supply an always-on
Researching Moltis? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Real-world workflow fit
Concrete scenarios for the personas Moltis actually fits — and what changes day-one when you adopt it.
Install with curl, open localhost:13131, sign in with GitHub Copilot OAuth, then ask the agent in natural language to run a recurring job that checks a repo's open PRs each morning.
Outcome: The cron job fires on schedule and posts its output back into the originating chat, with the agent remembering prior context across sessions.
Point Moltis at a local model via Ollama or LM Studio, store provider keys in the encryption-at-rest vault, and wire up Telegram plus Matrix as the only frontends used.
Outcome: Keys and conversation memory stay on hardware the operator owns, with sandboxed execution limiting what the agent can touch on the filesystem.
Register webhooks for GitHub PRs and Stripe payments, run the agent in Docker with sandboxed shell access, and route summaries into a Slack channel using per-message acknowledgment reactions.
Outcome: PR and payment events reach a single agent that can run sandboxed commands and report back in Slack without anyone watching a dashboard.
Use Cases
- Schedule recurring cron jobs in natural language and have the output land in the chat that created them.
- Browse and summarize web pages with SSRF protection and browser sessions isolated in containers.
- Manage your calendar and reminders through CalDAV and durable calendar/email connectors.
- Send and receive messages across Telegram, WhatsApp, Discord, Slack, Matrix, Nostr, and Teams from one agent.
- Extend the agent at runtime by creating custom skills, hooks, or MCP tool servers over stdio or HTTP/SSE.
- Automate workflows via webhooks, e.g. trigger on GitHub PRs or Stripe payments.
- Run a persistent local assistant that remembers context across sessions for research or personal tasks.
- Run sandboxed code from chat safely inside Docker, Podman, Apple Container, or WASM.
Models Under the Hood
as of 2026-10-18
Limitations
- Self-hosting means you manage the server, Docker/Podman, networking, vault unlock, and an always-on host.
- Moltis runs on macOS, Linux, or Windows hardware you own — there is no managed option in the sources reviewed.
- Every AI provider requires your own API key, an OAuth zero-config sign-in such as GitHub Copilot, or a local model you host.
- Provider and model availability changes: the September 2026 changelog removed an unavailable Fireworks model, so any pinned model choice may need replacing.
- The native iOS companion with APNS push is still listed as coming soon; PWA push notifications are the current mobile path.
- Configuration is TOML-based and the project deliberately has no third-party plugin marketplace, so extensibility comes from its own skills, hooks, and MCP tool servers.
as of 2026-10-09
Verification history
We have re-verified Moltis 9 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — re-checked, vendor evidence unchanged
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
Showing the 6 most recent of 9 verification passes.
Free to cite with attribution — this page re-verifies continuously.
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Plans compared
For each published Moltis tier: who it actually fits, and what it adds vs. the previous tier. Cross-reference the cost calculator above for projected annual outlay.
Self-hosted (open source)
$0/mo
Ideal for
Developers, privacy-focused operators, and small teams with an always-on Mac Mini, Raspberry Pi, VM, or dedicated server who are comfortable with TOML config and Docker/Podman.
What this tier adds
Starting tier — the software is free to self-host; your real spend is the host you provide plus your own provider API usage.
Where the pricing makes sense
The company stage and team size where Moltis's pricing actually pencils out — and where peers do it cheaper.
Moltis is open source and free to self-host; the real budget line is the always-on host plus your own provider API usage, so cost scales with how much you run rather than with seats. Compared with managed assistants that bundle hosting and model access into a subscription, Moltis shifts spend toward infrastructure and inference you control. It fits developers and small teams who already pay for a box and are comfortable with variable usage bills.
Setup time & first value
How long it actually takes to get something useful out of Moltis — broken out by persona, not the marketing-page minute.
Developers on a Mac Mini or Linux box: roughly 60 seconds from curl to a running server on localhost:13131, then minutes more to pick a model and connect a channel. Privacy-focused operators running local models via Ollama or LM Studio should budget extra time for model downloads. Teams wiring webhooks and multiple channels should plan an afternoon for TOML config, Docker or Podman setup, and
Switching to or from Moltis
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From OpenClaw: use the built-in read-only workspace import to bring over conversations, skills, and settings in one click.
- →From a managed cloud assistant: install the binary, paste or OAuth into your provider, and re-create prompts as natural-language cron jobs and skills.
- →From a local Ollama-only setup: point Moltis at the same Ollama endpoint and keep your GGUF models in place.
- ↗To OpenClaw: export sessions from the Web UI (Markdown copy and session export are built in), then re-import on the other side.
- ↗To a managed cloud assistant: keep exported session transcripts and reproduce scheduled jobs manually as platform-native automations.
Integrations
Resources & Guides
Tutorials & Learning
YouTube returned 6 videos for “Moltis”, and we withheld 6: 6 could not be judged, because “Moltis” is a single word that other videos use for other things. We are showing none, because we could not prove any of them are about Moltis.
Official links
Featured Head-to-Head Comparisons
Moltis vs Sublime Security
Moltis and Sublime Security serve entirely different needs. Moltis is a free, self-hosted AI agent for developers who want full control, multi-channel automation, and privacy. Sublime Security is a commercial email security platform for enterprise security teams battling advanced phishing and BEC. Choose Moltis if you need a programmable assistant; pick Sublime if you need to protect your organization from email attacks.
Moltis vs Audioeye
If you need a self-hosted AI agent that automates tasks across messaging channels and keeps data private, Moltis is the clear choice – it’s free and developer-friendly. If you must achieve web accessibility compliance quickly and reduce legal risk, AudioEye provides a managed turnkey solution with expert audits. They solve completely different problems, so the right pick depends on your priority: data control vs. compliance automation.
Moltis vs Push Security
If your priority is securing your organization against browser-based attacks and controlling employee AI tool usage, Push Security is the clear choice. If you need a self-hosted, private AI assistant for personal automation across messaging channels, Moltis delivers unmatched flexibility and data control. They solve fundamentally different problems—choose based on whether you're defending an enterprise or running your own agent.
Popular in Local & On-Device AI
Unsloth
Open-source local AI workspace for fine-tuning and running LLMs, image, and video models on your own GPU
Cortex.cpp
Run local HuggingFace models or route prompts to Claude, GPT and Gemini with your own API keys
Frequently Asked Questions
Best-of guides
Used Moltis? Help shape our editorial sentiment research.