What we're building for TAP on Telara
We've open sourced TAP at Telara so agents can build reusable internal tools. A primitive puts useful execution behind defined inputs, outputs and required access. Engineers can review, change or author that code as they would other internal tooling. Within a company, reuse raises some familiar engineering questions. Which version did we review? What assumptions does it make about the systems it…
We are developing TAP for Telara, an enterprise AI operating layer. TAP allows agents to create reusable internal tools by packaging useful execution behind defined inputs, outputs, and required access. When a team reuses this code, the interface must retain essential details such as the account, date range, source records, missing data, unmatched identifiers, and failed calls. This ensures the tool's results remain trustworthy.
When a reviewed package is executed, it operates on behalf of a specific individual or service. Different users may have varying levels of access based on company policy. The review of a procedure for shared use and the approval of changes during execution are distinct decisions. A company may approve a version for internal teams while still requiring authorization for any alterations requested by that version.
TAP utilizes the coding agent's existing tool connections, with Claude Code and Codex having the most extensive supported connected-tool paths. Other client paths have different compatibility limits. The standalone runtime is MIT licensed and does not necessitate a Telara account, service, or registry.
Telara serves as the enterprise AI operating layer, connecting AI clients to the company context they are entitled to use. Actions can run autonomously, require approval, or be blocked. Telara retains the resulting work and decision records, including tool calls, approvals, outcomes, and attributions. We are enhancing Telara by incorporating verification, versioning, admin approval, and distribution for TAP primitives.
By doing so, we aim to enable useful code created by one agent's tool to become reusable by colleagues' clients under company policy. Models, tools, and schemas will evolve, making reusable blocks essential for engineers and agents to inspect, test, and update, along with a record of their usage. Changes to a procedure's part should be reviewable without rebuilding the entire tooling.
I am part of the TAP team at Telara. If you oversee agents interacting with shared company systems, I would appreciate learning about your requirements for sharing an agent-written tool with another team, specifically regarding reviewing its version, authority, and failure behavior.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
