Agent Mode and Deep Thinking Mode are the two toggles inside PsyCortex that change what the engine does between your message and its reply. Agent Mode invokes tools and takes actions. Deep Thinking Mode reasons in private and returns only the clean, defensible final answer. Use Agent Mode when you need doing; use Deep Thinking when you need thinking.
- Agent Mode
- A PsyCortex mode in which the engine is allowed to invoke tools — web fetch, audit lookup, wallet top-up, deep research, file analysis — and chain multiple steps to complete a goal, not just a single message.
- Deep Thinking Mode
- A PsyCortex mode powered by PsyReason — Bharat NeuroTech's proprietary reasoning layer. A thinking status bar shows the engine is working; only the polished final answer is returned — internal reasoning stays private.
- Agent Mode = tool-use, actions, multi-step.
- Deep Thinking Mode = hidden reasoning, clean final answer.
- Both billed at the standard ₹0.50/msg meter — no premium tier.
- Toggleable inline via the chat header or slash commands
/agentand/deep. - Can be combined: agentic tool-use with private reasoning between steps.
When to pick which
| You want… | Pick | Why |
|---|---|---|
| Quick answer, no tools | Default | Cheapest, fastest |
| Reason through a decision | Deep Thinking | Hidden CoT, defensible final answer |
| Fetch data, run a workflow | Agent Mode | Tool-use, multi-step |
| Decode a hard question with research | Agent + Deep | Both toggles on |
How it fits the rest of the lab
Both modes use the same PsyCortex engine that powers the ₹1,799 audit chatbot and every custom deployment. The toggle is a runtime flag, not a separate model — so behaviour is consistent across chat, audit and bespoke instances.
Frequently asked
- What does Agent Mode actually do?
- It lets PsyCortex invoke tools — web fetch, audit lookup, wallet top-up, deep research, file analysis — and chain multiple steps to complete a goal rather than answer a single message.
- Why is the reasoning hidden in Deep Thinking Mode?
- To protect prompt design, avoid leaking partial answers, and give you a clean defensible final answer. A thinking status bar shows the engine is reasoning; only the final answer is shown and stored.
- Do the modes cost extra?
- No. Both modes are billed at the standard ₹0.50 per message, ₹1 per attachment meter — no premium tier.
- Can I use both modes at once?
- Yes. Agent + Deep combines tool-use with hidden reasoning between steps — useful for hard research questions that also need data fetches.
- How do I toggle them?
- Use the toggles in the chat header, or the inline slash commands /agent and /deep.
Try both modes — three free messages, no sign-up.
Open the chat, toggle Agent or Deep Thinking, and feel the difference on a real question of yours.
