Native Agentic Integration (MCP / Agent Skills) for Autonomous Agents like Hermes Agent
Maintainers usually reply within 1 day
Nobody has claimed this yet.
Assessment
- Difficulty
- 5/5
- Estimated time
- Over a week
- Newbie friendliness
- 25/100
- Issue type
- Feature
- Clarity
- Mostly clear
- Activity status
- Active
- Tech stack
- python, typescript
- Domain
- ai, backend-api-design
Research direction
Start by exploring the existing litellm integration and how Builder represents page trees, components, and custom scripts. The issue proposes an MCP server, expanded SKILL.md definitions, and an external agent backend, but does not identify files or tests. Done would require deciding the integration scope and providing a safe way for agents to inspect, edit, and validate Builder content.
Written by the indexing model from the issue text.
Description
Is your feature request related to a problem? Please describe.
Currently, AI features in Frappe Builder rely primarily on direct LLM calls via litellm (stateless single-turn generation). While this works for basic block generation, it limits complex multi-step workflows, iterative debugging, and full-stack site management.
In production self-hosted setups, many admins now use persistent, tool-equipped autonomous agents (such as Hermes Agent) to manage Ubuntu servers, ERPNext benches, and site maintenance.
Describe the solution you'd like
Instead of relying solely on embedded litellm calls for AI automation, it would be great if Frappe Builder and ERPNext embraced an agent-first architecture that external autonomous agents can plug into directly:
- Official MCP (Model Context Protocol) Server / Endpoints: Expose Builder page trees, components, custom scripts, and ERPNext DocTypes as structured MCP tools so agents like Hermes Agent can inspect, edit, and validate pages safely.
- Expanded Agent Skills (
skills/frappe-builder): Continue expanding standardizedSKILL.mddefinitions so external agents can natively understand Frappe Builder schemas, styling conventions, and deployment workflows without fragile scraping or raw file edits. - Configurable Agent Backend Option: Allow pointing Builder's AI assistant to an external agent endpoint (OpenAI-compatible API with session/tool support) rather than only raw model providers via
litellm.
Additional context
In our current Ubuntu + ERPNext setup, Hermes Agent already handles server/bench operations and troubleshooting reliably, whereas litellm inside Builder is isolated from the broader site context. Bridging this gap would turn Builder into a truly agent-ready web platform.
- Dominant language
- Python
- Stars
- 2.5k
- Forks
- 583
- Avg merge
- 19h 3m
- Merged PRs (30d)
- 87
Getting set up
Starts the project's dev container in your browser, under your own GitHub account.
- No Dockerfile or Docker Compose file
- No pull request template
- No contributing guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
More from frappe/builder
-
feature request
Difficulty 4/5 3-5 days Newbie friendliness 48/100
frappe/builder#1022 · 1 comment ·
Maintainers usually reply within 1 day
-
Analitycs optionsOpenfeature request
Difficulty 4/5 3-5 days Newbie friendliness 35/100
Maintainers usually reply within 1 day
-
bug released
Difficulty 3/5 1-2 days Newbie friendliness 35/100
frappe/builder#580 · 3 comments ·
Maintainers usually reply within 1 day
-
Difficulty 5/5 Over a week Newbie friendliness 35/100
frappe/builder#572 · 1 comment · 4 reactions ·
Maintainers usually reply within 1 day
-
Difficulty 3/5 1-2 days Newbie friendliness 30/100
frappe/builder#491 · 1 reaction ·
Maintainers usually reply within 1 day
Similar issues
-
needs-triage
Difficulty 1/5 1-3 hours Newbie friendliness 85/100
gke-labs/kube-agents#6620 ·
Maintainers usually reply within 1 day
-
bug
Difficulty 2/5 1-3 hours Newbie friendliness 75/100
open-webui/open-webui#34719 · 2 comments ·
Maintainers usually reply within 1 day
-
cognition
Difficulty 2/5 1-3 hours Newbie friendliness 82/100
Maintainers usually reply within 1 day
-
One non-object JSON record makes the session catalog exitPossibly taken @he-yufeng claimed this today. Open
Difficulty 2/5 1-3 hours Newbie friendliness 68/100
Maintainers usually reply within 1 day
-
Difficulty 1/5 Under an hour Newbie friendliness 78/100
collabnix/awesome-mcp-lists#433 ·
Maintainers usually reply within 4 days