Person

NAGARJUN DAS

GitHub@Nagarjundas1994-AiAgentslast push

Capabilities3

Provides 3
Agent toolingCAPAgent toolingcapabilityBuilding an agent into your own application — the tools it may call, the state it keeps between turns, and the framework it runs in. Distinct from MCP servers, which expose an application to somebody else's agent, and from agent guidance, which is knowledge handed to a coding agent.Input validationCAPInput validationcapabilityRejecting bad input before it reaches the database. While this can be done manually inside a handler, there are solutions in place to derive type, range, enum and mandatory checks from the model definitions.Tool governanceCAPTool governancecapabilityControlling what an agent may call, with which authorisations, and leaving an audit trail of what it did. The authorisation question that MCP exposure raises but does not answer.

Entries2

Group by
EntrySignals
Maintains2
cds-agentsv1.1.2
none

Toolkit turning a CAP service into tools for agent frameworks

cds-validatev1.0.0
none

Zod schemas as CAP validation middleware

Details

History3
  • cds-agentsdescription.what_it_solves, provides.agent-tooling → the producer-against-consumer line, the three entry points, the tool strategy and the discovery-cost argument, on a sourced edge with three property valuesThe entry stated the capability with no source, and its paragraph said the framework is the host without the fact that makes this entry different from every other option here: the README states it is a client library pointed at something already running, where @cap-js/mcp is a plugin whose application you must own. Read on 2026-10-15, which also gives the three entry points as the README scopes them, the toolStrategy values, the four-hundred-tool argument behind them, and the install line, which replaced an unsourced claim about four peer dependencies. The tool-governance edge is untouched: its claims belong to that capability's own pass.
  • cds-agentsprovides.agent-tooling.facets → brings: Tools for another framework and Loop in application code, harness: LangChain and LangGraph, protocols: OData V4capability:agent-tooling declared four comparison properties on 2026-10-15. Both brings values rest on their own sentence — CAPToolkit is the raw-tool layer and the production default, CAPAgent a ready-made ReAct agent for demos and small apps — and the protocol is the one its own executor names as a client, not a protocol it serves. The state and approval properties are left out: the README says nothing about either.
  • First entry seenDiscovered via npm registry · keywords:sap-cap, npm registry · keywords:cds.
For your agent
Ref
actor:nagarjundas1994-aiagents
JSON
/actors/nagarjundas1994-aiagents.json
MCP
Connecting an agent