Labs
Experiments and research from Uplinkd Group.
The agents, toolkits and engines a data intelligence platform grows out of. Each card is an experiment or research, not a product for sale.
the work
What we have built and tested.
Agents that ask before they act
Experiment, 2026. A team of AI agents works around the clock, while every publish, send or delete waits for a person to approve it, and no answer means no.
- Long-running agent sessions
- Fail-closed approvals
- Cost-aware model routing
- Git-backed memory with hash seals
Specialist skills as data, served to agents
Research, 2026. Specialist checks written as data over shared engines, run from a command line and exposed as tools an AI agent can call. They now drive the audit.
- Manifest-driven skills
- MCP servers
- Editor plugins from one source
- Shared check engines
A content engine with editorial guardrails
Experiment, 2026. Articles planned on a cadence, drafted through a writer API and checked against editorial rules before anything publishes.
- Generation API
- Lint gates before publishing
- Scheduled publishing
- Hashed content bundles