[Feature] Natural Language Agent Builder — "Describe Your Agent in Plain English"

[DevolRaman750]

🧠 Problem Statement

Creating an agent on AgentLog currently requires users to understand and configure:

• Heartbeat schedules
• Archetype templates
• Memory settings
• Token budgets
• Function selection (Slack, GitHub, etc.)

This is overwhelming for non-technical users — marketers, PMs, researchers, and
business users who want to use AI agents but don't know where to start.
The current setup creates a high drop-off rate at the agent creation step.

---

💡 Proposed Solution — Natural Language Agent Builder

Add a chat-style agent creation flow where the user simply describes what they want
in plain English, and the system intelligently configures the agent for them.

Example Flow:

User types:

"I want an agent that checks my GitHub repo every morning and posts a summary
of new issues to our Slack channel"

System automatically:

• ✅ Sets heartbeat to daily schedule (every morning)
• ✅ Enables GitHub + Slack function integrations
• ✅ Writes a sensible starter system prompt
• ✅ Sets a reasonable token budget (e.g. 5,000 tokens/day)
• ✅ Suggests an appropriate archetype template

User sees a preview card:
"Agent Name: GitHub Issue Watcher
Schedule: Daily at 8:00 AM
Functions: GitHub ✓ Slack ✓
Token Budget: 5,000 / day
System Prompt: "You are a GitHub monitoring assistant...""

User reviews, edits if needed, and clicks "Create Agent" — done in under 60 seconds.

---

🎯 Why This Matters

Without This Feature	With This Feature
Users need technical knowledge to configure agents	Anyone can create an agent by describing it
High drop-off at agent creation step	Faster time-to-value (< 60 seconds)
Only developers use the platform	Opens platform to non-technical users
Low activation rate for new signups	Higher retention and word-of-mouth growth

---

🏗️ Suggested Implementation

Frontend (React Native)

• New "Create with AI" button on the Agents page
• Chat input box: "Describe what you want your agent to do..."
• Live preview card showing auto-configured settings
• Edit mode — user can adjust any auto-filled field before creating

Backend (Go)

• New endpoint: POST /api/agents/generate-config
• Accepts: { "description": "user's plain english input" }
• Uses an LLM (via existing provider system) to extract:
  • schedule — heartbeat frequency
  • functions — which integrations to enable
  • system_prompt — starter prompt for the agent
  • token_budget — suggested daily limit
  • name — suggested agent name
• Returns structured agent config JSON

Prompt Design (for the LLM call)

"You are an AI agent configuration assistant.
Given a user's description of what they want an agent to do,
return a JSON config with: name, schedule_cron, functions[],
system_prompt, token_budget_daily.
Only return valid JSON. No explanation."

---

📋 Acceptance Criteria

• User can describe an agent in plain English (min 10 words)
• System returns a complete agent config in under 5 seconds
• User can review and edit all auto-filled fields before saving
• Agent is created and starts running on the configured schedule
• Works on Web
• Gracefully handles vague or unclear descriptions with a clarifying prompt
• Unit tests for the config generation endpoint

---

🔗 Related

• Extensibility docs in README: "Add providers: Implement the ModelProvider interface"
• Existing template system could serve as the output format for generated configs

---

🙋 Volunteering

I'm interested in working on this feature.
Would love feedback from maintainers on the proposed approach before I start coding.

@imran31415 f— happy to discuss the approach!

[imran31415]

Looks good! Let's do it

[DevolRaman750]

Sir , for this Natural Language Agent Builder I want to keep the system prompt and function read only .
Like No major schema/template editing will be done .This will not let the user edit the system prompt in preview and on save it will create a new custom template .
Is this approach right?