We've open sourced TAP so people and agents can build reusable internal tools. Once an agent has worked out a useful procedure, it can put the repeated steps in code and call that code with new inputs.
Within a company, the next question is who else can use it.
A customer-review procedure may collect CRM records, support incidents, billing data and product usage. A colleague might need the same process for another account. Reusing the code makes sense, but the colleague's agent still needs access to the right systems and records. The company needs to know which version was reviewed and what the procedure is allowed to do.
From one person's tool to shared company work
Telara is the enterprise AI operating layer. Connected AI clients receive company context they're entitled to use. Their actions run with the organization's identity, scope and policy: an action can run autonomously, require approval, or be blocked. Telara records the work it mediates, including tool calls, approvals, outcomes and who acted.
We're extending that platform with verification, versioning, admin approval and distribution for TAP primitives. We want a procedure created by one person's agent to become something colleagues' clients can use under company policy.
Reviewing a primitive for shared use and approving an action during execution are different decisions. A company might approve a version of a procedure for its teams while still requiring approval for a particular change it requests. Declaring access in a package does not grant that access; the host and the organization's policy still determine what the caller can do.
Keeping useful work
Models, tools and systems change. A useful internal tool should be something the team can inspect, test and update as its environment evolves. Smaller reusable blocks make it possible to change the part that needs attention and keep using the rest.
That's how TAP fits into our work on security, sovereignty and longevity for AI systems. The package describes its behavior and access needs. The open-source runtime can be used independently across compatible clients. Telara connects managed AI work to company authority and a retained operating record.
The standalone TAP runtime needs no Telara account, service or registry. We're building the team-management path so companies can review and share useful tools while continuing to choose the AI clients they use.
If your team is adopting AI across shared company systems, take a look at Telara. We're shaping this with enterprises and would like to hear how you want to review and share the tools your agents build.