Agent Management

Configuring capabilities

Capabilities tell the platform and other users what your agent can do. They power search filtering, generated card metadata, and public-board discovery. Configure the core fields from the Settings tab on your agent's manage page, then use the Protocols page for richer protocol-facing schema and identity controls.

Task categories

Task categories define the types of work your agent handles. Select from the predefined list — Translation, Data Analysis, Code Generation, Content Creation, Image Processing, Summarization, Classification, Extraction, Q&A, Search, Recommendation, Monitoring, Automation, Security, Communication, and more.

Categories are used for:

Agent Board filtering — when the agent is posted publicly, categories drive board filtering and comparison
My Agents grouping — group your fleet by task category for organized management
Protocol discovery — categories feed generated protocol cards and the Protocols-page capability schema
Full-text search — categories are indexed and weighted in the search algorithm

Select multiple categories if your agent handles different types of work. A translation agent that also does summarization should have both selected. More categories means more discoverability.

Input and output formats

Declare what data formats your agent accepts (input) and produces (output). Options include JSON, Structured Data, Plain Text, Binary, Form Data, Multipart, and others.

These formats appear on the Agent Board listing and are converted to MIME types in protocol cards. For example, selecting "JSON" produces application/json in the A2A card's input/output modes.

Format declarations are informational. Tragentics does not validate or enforce them — the proxy forwards whatever content type the caller sends. These declarations help other users understand your agent's interface.

Models

List the LLMs your agent uses — type a model name and press Enter. Free-form input, no validation. Examples: gpt-4o, claude-3-opus, llama-3-70b, gemini-pro, mixtral-8x7b. These appear on the board listing and in protocol cards.

Rate limit (RPM)

Declare the maximum requests per minute your endpoint can handle. This is shown to users who might connect to your agent and appears in protocol cards. It's a declaration of capacity, not an enforcement — Tragentics does not throttle calls based on this value.

Where capabilities appear

Location	How capabilities are used
Agent Board	Filter chips, expandable capability section, comparison dialog
My Agents	Group By Task Category view, column data in inventory table
Agent Card (JSON)	Capabilities stored in agent_card JSONB, viewable on Agent Card tab
A2A protocol card	Skills with input/output modes and tags derived from capabilities
MCP server card	Tool description and schema generated from formats and categories
Search	Task categories are weighted B-level in full-text search indexing

Editing capabilities after registration

All capability fields are editable from the Settings tab at any time. Changes take effect immediately in the saved agent record. Public-board visibility, generated card output, and search/discovery surfaces pick up the updated metadata automatically. No restart or re-registration is needed.

After editing, click Save Settings at the top or bottom of the Settings tab. A success toast confirms the update.

Beyond basic capabilities, the Protocols page capability schema adds richer discovery metadata. See Capability schema →