feat: add agent export and import functionality

[abhinav-m22]

/claim #4201

• Fixes Ability to Export and Import Agent configurations #4201

This PR introduces the ability to export an agent's configuration into a JSON file and import it into any other Archestra instance. This enables teams to move agents seamlessly across multi-environment setups (dev to staging to prod) and share configurations easily.

Features & Functionality

• Export Endpoint: GET /api/agents/:id/export generates a self-contained JSON schema representing the agent. Internal UUIDs are stripped out and replaced with portable names.
• Import Endpoint: POST /api/agents/import accepts the JSON payload, creates the agent, and intelligently resolves associations (Tools, Knowledge Bases, Connectors, Delegations) by name against the local registry.
• Strict Validation & Security: Association assignments (tools, KBs, delegations) strictly respect the importing user's RBAC visibility and team scoping.

Security Constraints & Edge Cases

• Import may return non-blocking warnings when tools/KBs/connectors/delegations can’t be resolved or assigned locally (e.g., remote MCP tools without required installations or credentials); the agent is still created, but those associations are skipped.
• Built-in Agents: Exporting built-in agents is explicitly forbidden (returns 400). They contain internal, non-portable configurations.
• Secret Exclusion: Security fields like identityProviderId and llmApiKeyId are intentionally excluded from the export payload because they contain instance-specific secrets.
• Imported agents are always forced to personal scope by default to prevent privilege escalation.

Demo showing both successful imports (no warnings) and imports with warnings for remote MCP tools:

Screencast.From.2026-05-01.17-12-37.mp4

[abhinav-m22]

Should we support importing MCP gateway and LLM proxy agent types, or restrict it to the standard agent type only? For now, this implementation restricts import/export to the agent type to keep v1 simple and stable.

[abhinav-m22]

Hi @joeyorlando 👋
Just wanted to kindly follow up on this PR as well. I completely understand things may be busy. Would really appreciate a review whenever feasible. Thanks 🙏

[joeyorlando]

hey @abhinav-m22 👋 thank you for the ping!

On first glance this makes sense. I will take a deeper review soon 🙂

My first comment would be that in the actions column for Agents, I think we should "simplify" that a bit - I would propose showing Connect, Chat, and Edit Icons followed by a "three dots" icon/button which opens a "pop-over" with the remaining actions. Does that make sense?