Can I add proof to AI work without replacing the tools my team already uses?
Yes. Proof is added as a layer, not a migration. Add the Zambo connection to the AI clients and frameworks your team already uses, and work run through it starts saving a verifiable receipt per step. Your models, prompts, repositories, and habits stay exactly where they are.
The compatibility question is the real buying question, because a proof tool that demands a new stack will be adopted by nobody with a working stack. The connection model answers it directly: Zambo is one hosted MCP endpoint, https://zambo.dev/mcp , added to the clients the team already opens (Claude, ChatGPT, Gemini, Cursor, Grok, Windsurf, VS Code) and reachable from the frameworks the team already builds with (LangChain, CrewAI, and the shipped AER-1 packages for agent frameworks).
What changes after the one-time addition is narrow and total at once. Narrow: nothing about how the team prompts, reviews, or ships. Total: every step that runs through the connection leaves a receipt with its result and source states, the day reads as one record across assistants, and delivery becomes a handover link instead of a summary the team writes by hand.
Name the boundary honestly, because your team will find it anyway: work that never touches the connection is not recorded. A chat in an unconnected client, a step run by hand, an agent framework without the package: those leave no receipts, and the record will show their absence as gaps. Adoption is therefore a routing habit, not a purchase event: point the client work at the connection, and let side experiments stay unrecorded by choice.
Checklist: the no-replacement adoption path
1. List the AI clients and frameworks the team actually uses for client work. 2. Add the connection to the top two; run one real job through each. 3. Read the combined record and check one receipt publicly. 4. Make the handover the default delivery for one friendly client. 5. Extend the connection to the remaining clients once the first delivery lands. 6. Write the one-line team rule: client work runs through the connection; experiments are free.
Worked example
Sample: a three-person team lives in Cursor for code, Claude for writing, and ChatGPT for research. Nothing is replaced. Each client gets the connection in a minute; the next client project's steps land in one record with three producer labels, and the delivery is one handover. The team's tools are untouched; their proof problem is gone. (Sample scenario; no live record is cited.)
Frequently asked questions
Does adding the connection change how the AI answers?
The connection adds tools and recording to the session; it does not retrain or restrict your models. Teams keep their prompts and their judgment. What they gain is that the work leaves a checkable trail.
We build agents with a framework. Is there a package, or only the hosted connection?
Both. Framework builders add the shipped AER-1 packages (for OpenAI Agents, AutoGen, LlamaIndex, Pydantic AI, and others) so their own agents mint receipts; the hosted connection serves people working in AI clients. Part 2 of this section carries the per-framework quickstart pages.
What is the smallest first step that proves compatibility?
One client, one real job, one checked receipt. If the record appears and recomputes, compatibility is proven for that client; repeat per client on the checklist. The demo gives the same proof in a minute with no setup at all.
By Brennan Zambo, founder of Zambo.
Last verified: 2026-10-17.
Try a handover
A handover gives your client one proof link with verifiable receipts to review, then Accept or Flag.
try it: say 'use zambo to hand this over to <client>' in your ai