Share Hindsight Memory Across ChatGPT and Perplexity
If you want to share Hindsight memory across ChatGPT and Perplexity, the key decision is not the connector itself. It is the bank boundary. Both tools can talk to Hindsight already, but they only feel like one workflow when they write to and read from the same memory bank on purpose. Use the ChatGPT guide, the Perplexity guide, the MCP server docs, and the quickstart guide as the technical references while you set up the shared path.
The quick answer
- To share memory, point both connectors at the same Hindsight bank instead of letting each tool keep its own isolated default.
- A shared bank works best when the work is genuinely shared, for example research in Perplexity and synthesis in ChatGPT.
- The fastest proof is to retain something in one tool, then ask the other tool a follow up question that depends on it.
Pick one bank on purpose
A shared memory setup is only useful if the bank boundary matches the work. Pick a bank name that maps to a real project, team, or ongoing research thread.
For example, if both tools are helping with the same launch, use one bank such as:
https://api.hindsight.vectorize.io/mcp/product-launch/
That keeps the context specific. If you pour unrelated work into the same shared bank, cross tool recall becomes noisy and the whole setup starts feeling less trustworthy.
Point both connectors at the same MCP path
After both tools are individually working, update each connector so the URL resolves to the same bank.
For example, both tools can target:
https://api.hindsight.vectorize.io/mcp/product-launch/
Then keep the custom instructions aligned. Ask both tools to retain durable findings, decisions, and constraints, not every casual aside. If ChatGPT stores architecture choices while Perplexity stores source backed research, the shared bank becomes much more valuable than either tool alone.
Test cross tool recall with one concrete workflow
A simple verification flow is enough:
- In Perplexity, research a topic and ask Hindsight to retain the key findings.
- Start a fresh ChatGPT chat.
- Ask ChatGPT to answer a planning question that depends on those findings.
- Check whether it recalls the earlier research instead of starting from zero.
You can reverse the test too. Store a decision in ChatGPT, then use Perplexity to continue researching from that decision state. The important part is not perfection on the first try. It is proving that both tools are reading from the same durable layer.
Use guardrails so the shared bank stays useful
A shared bank does not mean a sloppy bank. A few guardrails make a big difference:
- keep one bank per project or domain, not one bank for your entire digital life
- use consistent names for products, repos, and initiatives
- review noisy memories periodically in Hindsight Cloud
- split the work again if research and personal preference data should not mix
If you need deeper control, the next layer is the retain API, the recall API, and bank level configuration from the docs home. Shared memory is most useful when it stays scoped.
Where ChatGPT and Perplexity each help most
This setup works because the tools are different. Perplexity is good at web backed discovery, source gathering, and iterative research. ChatGPT is good at synthesis, planning, drafting, and reasoning through tradeoffs.
A shared Hindsight bank lets each tool leave behind durable work product for the other one. That is the real payoff. Instead of copying context across tabs, you let the memory layer carry it.
FAQ
Should both tools always share one bank?
No. Only do that when the work is genuinely shared. Separate banks are still the right answer for unrelated projects or different privacy boundaries.
Can I start with /mcp/default/ and switch later?
Yes. The important thing is consistency. Once you decide a workflow should be shared, point both connectors at the same named bank path.
What should I store from each tool?
Perplexity should usually retain research findings and sources. ChatGPT should usually retain decisions, constraints, drafts, and reasoning outcomes that matter later.