From team knowledge to shared intelligence.
Give your team and its AI tools a shared place to retrieve business knowledge. Start with a useful question, add the relevant sources, and control who can access them. Here is a practical first workflow.
- 01Choose a question
- 02Set access
- 03Add sources
- 04Check answers
- 05Connect AI
- 06Review access
Choose one useful question
Pick a question your team repeatedly answers, such as why a product decision was made. Start with one project and the documents that explain it. You do not need to connect a mailbox or configure an agent to begin.
Why did we choose a phased rollout for this release?
A specific question helps you judge whether the retrieved context is useful.
Set the access scope
In Access, create a scope for the project and review its grants. Grants can cover nested scopes, so check inherited access too. Choose the matching scope in Memory before uploading material. Only share content you are allowed to use.
Project knowledge for the people working on that project.
Scopes organize access. A grant defines who can read or write within that boundary.
Add the source material
Open Memory and use Upload to add a document, transcript or text file. Wait for processing to finish before asking about it. Keep the original source available so you can compare extracted knowledge with what was actually written.
A product brief, a decision note and a research transcript.
Start with a small set of relevant sources, then add context when it answers a real need.
Ask and check the evidence
Use Memory → Ask to ask your question. Explore the graph to inspect related knowledge. Review the available source context, dates and missing information before acting on an answer. Retrieval can miss relevant material, and generated answers can be wrong.
What supports this decision, and what is still uncertain?
Compare the answer with your source documents. A fluent answer is not proof of accuracy.
Bring your own AI when useful
Settings → MCP provides a hosted connection and client configuration. Create a token with the scope, permissions and expiry needed for the task. Use a client that supports the supplied connection method. Built in agents are optional.
https://willder.ai/api/mcp
Use the configuration shown in your workspace. Client support varies; keep the token private.
Review access as the team changes
Use Access to review grants and tokens, revoke access that is no longer needed, and inspect audit activity. Review both user grants and agent credentials when someone leaves. Revocation limits future requests; it cannot erase information already copied elsewhere.
A contractor finishes. Review their grants and any tokens issued for their work.
Memory and access need maintenance as the project and the people working on it change.
A few things worth understanding
Is shared AI memory the same as chat history?
Chat history records a conversation. Shared memory lets authorized people and tools retrieve knowledge across uploaded sources and workflows. It still needs review for missing or outdated information.
Do I need to run Willder agents?
No. You can start with Memory → Ask and Explore. Add an MCP connection or configure an agent when a repeatable task makes it useful.
Who can read the memory?
Access depends on the caller's permissions, scope grants and any token restrictions. Review inherited grants as well as direct grants before sharing project knowledge.
Does connecting an AI make my memory public?
An authorized connection can retrieve information within its permissions. Treat anything sent to that client according to the client's own data policies. Public search discovery concerns this website, not access to your workspace.
Does revocation delete old answers?
No. Revoking a grant or token can stop future authorized requests through that access path. It cannot retract data already returned to a person or another system.
Does a memory graph guarantee better answers?
No. Results depend on the sources, retrieval and the question. Try a few real team questions and compare the answers with the original documents before relying on them.
Ready to run it.
Bring one team question and the sources behind it. We can help you explore whether shared memory fits your workflow.
Talk to sales