Raposa vs LangChain / LangGraph
Both let a human approve an agent action. They sit at different layers: LangGraph pauses inside your graph; Raposa is a framework-agnostic approval layer with a verifiable audit trail. Here is the honest difference.
The short version
LangGraph — the graph runtime in the LangChain ecosystem — has first-class human-in-the-loop support. You can interrupt() a graph, persist its state through a checkpointer, and resume once a human provides input. It is a strong fit when you are already building on LangGraph and want to pause a run for review inside your own application, owning the persistence, the reviewer UI and any record you keep.
Raposa is not an agent framework. It is a dedicated approval layer any agent calls over HTTP — including a LangGraph node. It adds the parts you would otherwise build yourself: signed approval channels (email, Telegram, Slack), named approver groups with N-of-M sign-off, reminders and escalation, and a hash-chained, exportable audit trail. The two are complementary as often as they are alternatives.
Side by side
LangChain / LangGraph details reflect its public documentation as of September 2026. Where its docs leave something to the developer, we write "you build it" rather than imply it cannot be done.
| Dimension | Raposa Aval | LangChain / LangGraph |
|---|---|---|
| Category | Framework-agnostic human approval layer (HTTP API) | Agent framework with a built-in human-in-the-loop interrupt |
| HITL approval scope | One API call gates any action from any framework | Pause/resume of a LangGraph run inside your app |
| Where the pause state lives | Managed by Raposa | Your checkpointer (in-memory, SQLite, Postgres, etc.) |
| Approval channels | Signed email links, Telegram card, Slack message — button decides | You build the reviewer UI / channel |
| Named groups & N-of-M sign-off | Built in, every signer recorded | You build it in application logic |
| Audit trail verifiability | Hash-chained, signed, exportable, tamper-evident | Graph state / your logs; not a cryptographic audit artifact |
| Reminders, escalation, expiry | Built in and audited | You build it |
| Deploy / payment / comms gating | Same one call for any high-stakes step | Any step you code an interrupt around |
| Hosting & data location | EU only (Germany), DPA available | Wherever you self-host, or LangGraph Platform region |
| Framework coupling | None — call from LangGraph, OpenAI SDK, n8n, custom | LangChain / LangGraph ecosystem |
When to choose which
Choose LangGraph's built-in HITL if you are committed to the LangChain ecosystem, the human step lives entirely inside one graph, and you are happy to own the persistence, the reviewer interface and whatever record you keep. It is a clean, native way to pause a run for input.
Choose Raposa — or add it inside a LangGraph node — if you need approval channels people actually use from a phone, named groups with two-signature sign-off, and above all a verifiable audit trail you can hand to an auditor, ideally hosted in the EU. It also keeps working unchanged if you later move off LangChain.
Add a provable approval layer to your LangChain agent
Call Raposa from a LangGraph node with one HTTP request. Start free — 100 approvals a month, full audit export, no card.