Agent

Also for this capability: API · CLI · MCP · SDKs

agent: 38 operations. Name one in "op" and pass that operation's own arguments in "input". describe returns an operation's input schema.

Arguments

op — what each one does:

• delete_agent — Removes an agent and every run recorded against it.
• delete_agent_target — Deregisters one machine.
• list_agents — Returns every agent defined in the caller's org, each with the number of runs recorded against it.
• get_agent_activity — Serves the org-wide recent-activity feed.
• list_agent_builds — Returns the public index of every published build, most recently updated first, so a gallery can link straight to the story behind each product.
• get_agent_build — Returns the readable build of one product: the agent session that produced it, turn by turn — the prompts, the reasoning, the commits each turn produced — plus the exact git log that re-derives every commit binding from git itself, so nothing here has to be taken on trust.
• get_agent_by_ref — Returns one agent with its system prompt and its 20 most recent runs.
• get_agent_by_ref_runs — Returns one agent's execution history, newest first — each run's input, its output or its error, and how long it took.
• get_agent_spend — Answers what one of your org's agents has spent, in integer micro-USD.
• list_agent_coding_artifacts — Lists what a coding run left, kept after its sandbox is gone: every file it added or changed and its whole change as changes.patch, stored beside the run; the ports it served, each a preview while its sandbox is kept; its pull request and where its work was published.
• get_agent_coding_blob — Returns one file of a coding run's repository, read where the tree is read, in the shape GET /v1/git/repos/{name}/blob answers: text verbatim, anything else base64, and a file past the 1 MiB view cap marked truncated with no content.
• list_agent_coding_changes — Returns what a coding run changed, read from the forge it pushed its branch to: the commits on its branch that the base does not have, newest first; the net change of the branch against its base, one entry per file with that file's patch; and its pull request with the reviews it has had, or null while it has none.
• get_agent_coding_tree — Lists one directory of a coding run's repository, one level down with directories first: at the run's own branch once the forge holds it, and at the branch it started from until then — ref says which.
• list_agent_metrics — Serves the invocations-over-time histogram for the org's Agents dashboard.
• get_agent_runs — Returns the org's agent runs across EVERY agent, newest first — what ran here, for whom, on which model, how long it took, and why it failed.
• list_agent_sessions — Returns the caller org's live sessions, newest first — each with its event count, its direct-child count and a one-line preview of its latest event.
• get_agent_session — Returns one session with its direct child sessions and its 50 most recent events, oldest of those first.
• get_agent_session_control — Returns the steering commands (pause/resume/stop/message) recorded against the caller's own session that are newer than the cursor, oldest first, with the cursor to poll from next.
• get_agent_session_progress — Returns how far along one run is: the share of its goal that is done, whether it is running, blocked or finished, and a line saying what it is doing right now.
• get_agent_session_tree — Returns the subagent-flow graph rooted at this session: the session, its children, their children, each node carrying its own event count.
• list_agent_targets — Returns every machine registered to the caller's org, newest first, each with its live session load.
• get_agent_target — Returns one registered machine, with its live session load.
• update_agent — Changes an agent in place.
• update_agent_session — Updates a session's surface-owned truth: its status, its title, the run-target it is dispatched to, and the product it built plus whether that build's story is public.
• update_agent_target — Updates one machine in place.
• create_agent — Defines an agent in the caller's org: a model, a system prompt (instructions) and a set of tool names.
• create_agent_coding — Start one autonomous coding run against a repo in the caller's org
• merge_agent_coding — Merges a coding run's pull request into the branch it proposes into, on the forge the run pushed to, and answers the pull request after.
• create_agent_session — Opens a live agent session in the caller's org — the row every surface (the CLI's outer agent, hanzo.bot, the console, chat) hangs its activity off.
• budget_agent_session — Sets, raises, or removes a session's cap.
• create_agent_session_event — Records one turn of a session's transcript and answers 201 with it.
• message_agent_session — Sends a steering message to a running session — the endpoint a human or another agent interrupts through.
• pause_agent_session — Asks a running session to pause.
• resume_agent_session — Asks a paused session to continue, on the same terms as a pause.
• stop_agent_session — Ends a running session.
• create_agent_target — Registers a machine as an agent target, or re-links one that is already registered.
• claim_agent_target — The machine's long poll for work: it authenticates the daemon, stamps the liveness the dispatch gate reads (the poll IS the proof a runner is listening), and waits up to 25 seconds for the next run addressed to THIS machine.
• report_agent_target_run — Completes a claimed run: it delivers the terminal result to the run's durable owner, which is what lets that workflow finish.

tools/list declares a type, a description, which fields are required and an enumerated value set for each field and nothing further, and every column above is MCP's own. This tool takes an operation name and that operation's arguments, so what the document constrains is what goes inside input, field by field, on the operation you name — ask describe for that. A — means MCP does not constrain the field.

Call it

A tools/call carries every argument in one flat object — nothing binds to a path or a query string. This call carries exactly the arguments the operation requires, so it is the smallest one that can run.

curl -X POST https://api.hanzo.ai/v1/mcp \
  -H "Authorization: Bearer $HANZO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
       "jsonrpc": "2.0",
       "id": 1,
       "method": "tools/call",
       "params": {
         "name": "agent",
         "arguments": {
           "op": "list_agents"
         }
       }
     }'

Values are the operation's own defaults and enumerated values where it declares them, and a <placeholder> where neither source declares one. tools/list needs no credential; tools/call does — called without one MCP answers HTTP 200 with a JSON-RPC result whose isError is set and whose text says what was missing. How to get a key →

The operation behind it

agent dispatches to 38 operations. 4 of them MCP names exactly as the document ids them, and those are below; for the rest MCP has its own verb, which describe resolves.

The same capability over plain HTTP is in the agent API reference, on https://api.hanzo.ai.

All 120 tools · MCP · API reference

Generated from tools/list on https://api.hanzo.ai/v1/mcp — 120 tools captured 2026-09-30.