Artificial Wit vs Nova

Artificial Wit and Nova are both ai agent apps tracked by AIDiveForge. Below is a side-by-side comparison of pricing, capabilities, platforms, and ownership — sourced from each tool's live website and verified before publishing.

Artificial Wit

The platform sits between your existing APIs, documents, and knowledge bases on one side and any LLM — Claude, ChatGPT, Gemini, or a local model — on the other. You connect REST or GraphQL endpoints, upload docs or point at a database, then the platform exposes every configured API as a Model Context Protocol tool, discoverable by any MCP-compatible client. No schema migration, no re-platforming. The free tier caps you at three API connections, which covers a proof of concept but hits the wall fast for a real ERP environment. Role-based access control is included, which matters the moment clinical documents or order data enter the picture.

Nova

Nova runs 24 specialist agents on your own machine against your own API keys, storing everything in local SQLite. The pipeline is explicit — classify, decompose, prepare, then stop at a gate before anything publishes, sends, or spends. That gate is the differentiator. Where it strains: the roster of 24 agents with no hosted fallback means setup requires real technical lift, and the Bun + TypeScript stack narrows who can extend it. Teams that outgrow the built-in connectors or need agents to reason across domains the roster doesn't cover will find themselves writing custom system prompts before they expected to.

Attribute	Artificial Wit	Nova
Pricing	Paid	Free
Free trial	No	No
Open source	No	Yes
Has API	Yes	No
Self-hosted option	No	Yes
Platforms	Web, MCP clients (Claude Desktop, Cursor, ChatGPT)	Local machine, VPS, terminal, Telegram, Slack, Discord
Pros	No-code MCP tool generation from any configured API, which means Claude Desktop or Cursor can call your internal ERP endpoints without a custom integration build for each LLM client. Retrieval-augmented generation with cited answers baked into the knowledge base, so the assistant returns sourced responses from your actual documents rather than the model's training data — which removes the audit problem for healthcare and compliance contexts. Role-based access control included at the platform level, so permission-aware queries against clinical documents or financial data do not require a separate access layer bolted on afterward. Provider-agnostic LLM routing — OpenAI, Anthropic, Gemini, or a local model — so swapping the underlying model when pricing or performance shifts is a configuration change, not a rebuild. Agents coordinate multiple APIs and knowledge sources in a single workflow, so a query that needs to cross-reference an ERP record and an internal policy document does not require you to wire those calls together manually.	Two-phase execution with an explicit approval gate per action category, which means agents that connect to live ad accounts or email cannot publish or spend without you signing off — eliminating the class of 'it already fired' incidents that burn trust in agent tooling. All data — messages, memory, tasks, embeddings — lives in local SQLite computed on your own machine, so teams under data residency requirements or handling sensitive credentials get full perimeter control without building a custom storage layer. Provider-agnostic routing across Claude, Gemini, and Codex CLIs under your own API keys, with rate-limit fallback, so a provider outage or cost spike doesn't halt the operation — you change the route, not the architecture. Playbooks let you author a standard operating procedure once and re-run it with variables, so repeatable multi-step work (campaign launches, onboarding flows) doesn't require a human to re-orchestrate every instance. MIT license with full self-hosted deployment, which means there is no vendor dependency to negotiate and no pricing gate between you and the source code when you need to audit or modify agent behavior.
Cons	The free tier limits you to three API connections — a single ERP integration with a knowledge base and one additional service exhausts it. Teams scoping a real enterprise deployment hit the paid tier before the pilot is done. No self-hosted deployment option is available. Organizations under data-residency mandates, HIPAA BAA requirements that prohibit cloud egress, or air-gap security policies cannot use the platform at all — those teams evaluate self-hostable alternatives instead. Agent configuration happens through a no-code admin panel, which covers straightforward tool-call chains but gives no indication of supporting complex branching logic based on intermediate results. Teams that need multi-step conditional flows — branching on what the ERP returned before deciding which knowledge base to query — report adding a custom orchestration layer, at which point they are maintaining two systems. The platform is not open-source, so debugging unexpected agent behavior or auditing how credentials are handled requires trusting vendor documentation rather than inspecting the runtime directly. For security-sensitive enterprise procurement, that gap extends review cycles.	The self-hosted-only model requires a working Bun + TypeScript environment before a single agent runs — teams without a developer who can own that setup will stall at installation, not at agent logic, and the absence of a hosted option means there is no fallback path. The 24-agent roster is the ceiling on built-in specialization: when a workflow needs a domain the roster doesn't cover, extending Nova means writing and maintaining custom system prompts with their own tooling and skills — at which point you are building an agent, not using one. Durable multi-day processes that wait on timers or external signatures depend on the self-hosted instance staying live; teams without reliable server infrastructure will lose in-flight processes on restart, and there is no managed persistence layer to absorb that failure. Teams that need a no-code interface for non-technical operators configuring automations will hit a wall immediately — the interface is chat-plus-terminal, and authoring playbooks or wiring connectors requires direct file or config editing; at that point they are evaluating hosted agent platforms with visual builders instead.

Bottom line

Artificial Wit is paid while Nova is free; Nova is open source; only Artificial Wit exposes a public API. Choose based on which difference matters most for your workflow.

Frequently asked questions

What is the difference between Artificial Wit and Nova?

Artificial Wit is Paid, while Nova is Free and open source. Compare pricing, free trial, API, platforms, and pros/cons in the table above on AIDiveForge.

Is Artificial Wit better than Nova?

It depends on your workflow. Use the side-by-side attributes (pricing, open source, API, self-hosted, platforms) to decide. AIDiveForge does not rank a universal winner — we publish verified facts so you can choose.