OpenTag is an AI coworker that lives inside Slack, turning threads into finished work by investigating issues, updating tools, and running recurring jobs. It is model-agnostic, routing each task to the best-suited AI model while maintaining per-person permissions and requiring human approval before any irreversible action. OpenTag also builds a living company wiki from its work, preserving institutional knowledge and keeping context accessible.
Funding
Funding not disclosed
Founders
Product
Problem
Teams lose time and context when they move work out of Slack into separate AI tools, forcing them to rebuild the surrounding decisions, people, and history before any useful work can start. Important company knowledge is scattered across channels nobody searches, so answers and recurring workflows depend on the people already carrying that context.
Solution
OpenTag is a model-agnostic AI coworker that lives directly in Slack, where work already happens. Users tag OpenTag in a thread to investigate issues, update tools, or run recurring jobs, and it reports finished work back in the same conversation. It connects to approved systems, chooses the best AI model for each step, and asks for approval before sending, spending, or submitting anything. OpenTag also reads the company's channels to build a living wiki that captures decisions, ownership, and repeating workflows, so context compounds instead of walking out the door.
Target Audience
Primary users are engineering, operations, and growth teams at startups and small companies that rely on Slack for daily coordination and want an AI coworker that can act on their tools while keeping human oversight.
Features
- Slack-native interface: tag OpenTag in any thread to delegate tasks and receive results in the same conversation
- Model-agnostic routing: each step is sent to the best-suited AI model, keeping workflows stable while improving performance and lowering costs
- Proactive job ownership: after spotting a repeated task, OpenTag volunteers to own it, such as running the Monday numbers every week at 9:00 and posting to #growth
- Per-person permissions: integrations are scoped to the person who connected them, and a read-only tool cannot write at all
- Approval gates: OpenTag waits for human sign-off before sending, spending, submitting, or purchasing, and every action is audited
- Living company wiki: OpenTag reads invited channels and maintains a current wiki with sources, so answers exist before anyone asks
- Source-cited work: OpenTag names the model used, shows its plan, and cites sources so mistakes stay reversible