opentag

Tag an agent from anywhere. Stream it to everyone. Deliver it anywhere.

Download · Report Bug · Go Docs

     

Status

opentag is beta software. It is pre-1.0 and not yet recommended for production.

Interfaces are stabilizing but may change between minor versions, and a minor release may require code changes on your side.

What that means concretely today:

• Works and is tested. The single-binary deployment (in-memory bus, ledger and registry), the topic and delivery model, durable runs with replay, the Connect and SSE transports, and the Slack, GitHub, Jira, cron and webhook connectors.
• Written but not exercised in CI. The Redis Streams bus backend passes the shared conformance suite, but its integration tests skip without a Redis to point at.
• Not built yet. A persistent registry: agent specs live in memory, so they do not survive a restart.

See Limitations for the full list.

opentag is an agent platform and a low-latency pub/sub bus. Define an agent with a spec, tag it from Slack, GitHub, Jira or a schedule, and its events stream to any subscriber while being delivered to any destination you name.

The idea it turns on: where a tag is raised and where its answer lands are independent. A review request raised in GitHub can answer into Jira. A nightly schedule nobody watched can answer into both. That is what makes it a mesh rather than a pile of integrations.

Features

• Agents are documents. Create one by submitting a spec. No redeploy, no code, no restart.
• Listen to an agent. One topic namespace, matched by prefix: agent:docs-bot receives every run of that agent, including runs that start after you subscribe.
• Origin and destination are independent. Any trigger can raise a tag; any sink can receive it. They never call each other, they only publish and subscribe.
• N:M delivery with per-route selectivity. One run streams every delta into a GitHub comment it keeps editing while a webhook on the same run receives only lifecycle.completed.
• Connectors act, not just talk. The Actor face contributes native verbs (transition a Jira ticket, request a review) as agent tools, sandboxed by NGAC, each emitting an auditable action.taken event.
• Durable by construction. Runs survive crashes and replay from a ledger. A redelivered webhook joins the run it duplicated instead of answering twice.
• Revisions are pinned per run. A run accepted under revision 6 executes as revision 6 forever, even after a revise, because replay demands it.
• Citations are first-class events, carrying the exact quoted source and resolvable provenance.
• One contract, three transports. Connect serves gRPC, gRPC-Web and HTTP/JSON, plus resumable SSE for browsers.

Installation

Script (macOS / Linux)

curl -fsSL https://raw.githubusercontent.com/urmzd/opentag/main/install.sh | sh

Go

go install github.com/urmzd/opentag/cmd/opentag@latest

Library

go get github.com/urmzd/opentag

Quick Start

See the whole architecture run in one process, with no API key, no network and no infrastructure:

go run ./examples/local

It creates an agent, revises it, tags it from a simulated GitHub trigger with delivery routed to Slack and a webhook, shows a third party listening to the agent's topic, and demonstrates that a replayed run does not pay for the model twice.

Or run the real thing:

opentag serve &

cat > docs-bot.json <<'JSON'
{
  "name": "docs-bot",
  "description": "Answers questions about the deploy pipeline",
  "provider": "anthropic",
  "model": "claude-sonnet-5",
  "system_prompt": "You answer questions about the deploy pipeline. Cite your sources."
}
JSON

opentag agent create -f docs-bot.json
opentag tag docs-bot "how does a release get out the door?"

Deliver the answer somewhere other than your terminal, with each destination choosing its own granularity:

opentag tag review-bot "review this change" \
  --deliver 'github://urmzd/opentag/issues/42|delta' \
  --deliver 'webhook://acme/deploys|lifecycle.completed'

Watch the bus. This subscriber invoked nothing and still sees every run:

opentag listen agent:docs-bot
opentag listen agent --kinds lifecycle.completed --format json | jq

Commands

Command	Purpose
opentag serve	Run the core: API, SSE, webhooks, bus, router, worker
opentag work	Run an execution node with no inbound API
opentag agent	Create, revise, get, list, history, delete
opentag tag	Tag an agent and stream its answer
opentag listen	Subscribe to a topic and print events
opentag version	Version, commit, build date
opentag update	Self-update from GitHub releases

Every command takes --format text|json. Results go to stdout, diagnostics to stderr, so --format json output pipes cleanly into jq even while the same process logs a reconnect.

Documentation

Document	Covers
docs/architecture/overview.md	Topics, connector faces, N:M delivery, revision pinning, package DAG, limitations
AGENTS.md	AI-facing conventions, commands, rules, extension guide
examples/README.md	Runnable example index
CONTRIBUTING.md	Development workflow and commit convention
SECURITY.md	Security boundaries and reporting

API reference: pkg.go.dev/github.com/urmzd/opentag

How it composes

opentag builds on three libraries and adds addressing, distribution and the control plane:

Concern	Owner
Agent loop, typed deltas, RAG, citations	saige
Durability, replay, per-run journal	duraturo
Sandboxed execution nodes, NGAC	dispatch
Topics, bus, routing, connectors, specs	opentag

Limitations

• The bus requires Redis or the in-memory backend. duraturo's pgqueue does not implement DeltaLog, so a Postgres-only deployment runs durably but cannot stream.
• The registry is in-memory only; a Postgres implementation is not written yet.
• dispatch's queue is at-most-once in beta, so durability comes from duraturo rather than from dispatch.
• Citations are derived from RAG context blocks, because saige v0.14.0 has no agent-level citation delta.
• The bus is at-least-once with no consumer groups. If durable competing consumers become a requirement, NATS JetStream behind the Bus interface is a better answer than growing a broker here.
• Agent specs load from JSON, not YAML.

License

Apache 2.0. See LICENSE.