.png?width=160)
Bevel
Understanding legacy code through docs and diagrams
FreeDocs319 impressions#532 of its week
Bevel provides a vendor-agnostic, git-backed control plane to define, version and govern enterprise AI agents, their context, skills and tool access.
- for
- Enterprise developers and ops teams building and managing AI agents at scale
- pricing
- open source
- license
- open source
Key features
- Typed Knowledge Nodes — Store facts with provenance in a traversable graph for dashboards and mass updates.
- Markdown Skills — Define agent procedures in plain Markdown, reviewable via diffs.
- Tool Manifests & Permissions — Declare tools and access rules in YAML, with secrets stored in a vault.
- Per-Agent Identity — Each agent has its own credentials and scoped permissions.
- Git-Backed Configuration — All definitions live in Git, using branches, change requests and diffs.
- UTCP Protocol & CodeMode — Open universal tool-calling protocol and library for chaining tools without a proxy.
Use cases
- Version-controlled AI agent definitions across multiple runtimes
- Securely expose internal APIs to agents via tool manifests
- Govern tool access and audit actions per agent
- Integrate existing enterprise tools (e.g., Salesforce, SharePoint) into AI workflows
- Build reusable AI agents for RFI handling, GTM campaigns, market intelligence
Bevel vs alternatives
Bevel | Mintlify | GitBook | Moxie Docs | Outline | |
|---|---|---|---|---|---|
| Best for | Unified control plane for AI agents | AI-agnostic documentation platform | Team knowledge base | Living docs tied to GitHub repos | Quick collaborative knowledge base |
| Pricing | Open source | Subscription | Subscription | Subscription | Free |
| DevHunt upvotes | 0 | 0 | 0 | 4 | 0 |
| Launched | Jan 2026 | — | — | Jul 2026 | — |
- Bevel vs Mintlify: Focuses on generating developer documentation, not on AI agent governance.
- Bevel vs GitBook: Provides a docs knowledge base, lacking control-plane features for agents.
- Bevel vs Moxie Docs: Offers living docs with MCP context but does not include full vendor-agnostic control plane or UTCP.
- Bevel vs Outline: Fast team wiki without AI agent management or tool-calling protocols.
Bevel FAQ
What is the core of Bevel and can I use it for free?+
The core, called Hexis, is open source and available on GitHub at no cost.
How are agent definitions stored and managed?+
All context, skills, tools and access policies are kept in plain Markdown/YAML files in a Git repository, using standard Git workflows.
Can Bevel work with my existing AI runtimes?+
Yes, Bevel serves definitions to any agent runtime, including Claude Code, Cursor, ChatGPT, opencode and self-hosted models.
What does the Universal Tool Calling Protocol (UTCP) provide?+
UTCP lets you describe existing tools as HTTP endpoints so agents can call them directly without a proxy, enabling scalable multi-tenant AI applications.
Is there a way to chain multiple tool calls efficiently?+
The CodeMode library enables tool-chaining through code execution rather than single sequential calls.
How does Bevel handle secrets and permissions?+
Secrets are stored in a vault and access rules are defined per-agent in policy files, with changes reviewed via Git change requests.
Summarized by DevHunt from bevel.software · Oct 6, 2026. Details may change; check the official site.
Bevel
Mintlify
GitBook
Moxie Docs
Outline








