Version · publish · run

Build an agent once, call it from anywhere.

Author a prompt, an agent, or an image project, iterate in versions, publish one — then call it from the console, an OpenAI-compatible API, Slack, a webhook, or another agent. Every run attributed, priced, and bounded.

Your Google account, on one of this deployment’s allowed domains.

One engine

Every model, every surface

Live traces

Every agent handoff

Exact cost

Every call attributed

agent run · text/event-stream

live

POST /api/projects/support-triage/versions/published/agent
data: {"delta":{"content":"Looking at the report…"}}
data: {"delta":{"toolCalls":[{"id":"call_1","type":"function","function":{"name":"Skill","arguments":"{\"skill_name\":\"triage-rules\"}"}}]}}
data: {"toolResult":{"toolCallId":"call_1","name":"Skill: triage-rules","content":"# Triage rules…"}}
data: {"delta":{"toolCalls":[{"id":"call_2","type":"function","function":{"name":"transfer_to_agent","arguments":"{\"agent_name\":\"escalation-agent\",\"message\":\"Rate this crash report.\"}"}}]}}
data: {"author":"escalation-agent","authorPath":["escalation-agent"],"delta":{"content":"Severity: P2"}}
data: {"author":"escalation-agent","authorPath":["escalation-agent"],"authorDone":true}
data: {"toolResult":{"toolCallId":"call_2","name":"transfer_to_agent: escalation-agent","content":"Transferred to 'escalation-agent'; its answer follows.","displayOnly":true}}
data: {"delta":{"content":"Filed as P2 with repro steps."}}
data: {"usage":{"inputTokens":812,"outputTokens":164,"costUsd":0.0031}}
data: {"done":true}
data: [DONE]

● agent online

What AgentDure covers

projects

Projects & versions

Author prompts, agents, and image projects as immutable versions. Publish one; callers pin it or follow the pointer.

agent

Agent loop

A multi-turn tool loop with turn and context budgets, on-demand skills, subagent transfers, and opt-in URL reading — streamed end to end.

mcp

MCP tools

Register a server once; versions bind it, narrow its tools, and override headers — with per-project OAuth, secrets encrypted at rest.

skills

Skills

Markdown behavior packs, listed to the model and loaded only when it asks.

plugins

Agent Plugins

Skills and MCP servers sync from one plugins repo — the source of truth for every name it declares.

catalog

Capability catalog

One searchable index of skills, MCP tools, and external agents. A version that opts in has its bindings widened per run by what the prompt and the request ask for.

chats

Chats

Talk to any agent project — attach images and documents, replies stream, tool traffic stays inline, and a run outlives the tab that started it.

images

Images

Draw or edit from a prompt — as a project type, agent builtins, or an image subagent; an edit can address any image the run has seen.

artifacts

Artifacts

Every image or file a run produced, stored under a signed address and listed per project and per person — whichever surface it came from.

surfaces

Slack, A2A & webhooks

Per-project Slack bots that answer mentions, DMs, and keyword-matched channel messages, A2A in both directions, webhook and schedule triggers — every entry point runs the same engine.

cost

Cost & guards

Every call priced — what the channel charged, or the model registry's rate — and rolled up per project, per caller, per day. Daily and monthly thresholds warn, then refuse; concurrency and member tiers bound the rest.

traces

Traces & audit

Every agent run traced turn by turn, tool traffic included, and reachable from its project. Secret reveals, admin overrides, and deletions each leave an audit row.

Dure (두레) — a Korean village work cooperative, where neighbors pool their labor to finish what no one could alone. Agents here work the same way.

An internal LLM platform for prompt, agent, and cost management.

v0.70.2