Split.io (Harness Feature Flags)
Feature flag and experimentation platform (now part of Harness) with REST API and SDKs for managing feature splits, measuring impact with metrics, and safely rolling out changes.
Score Breakdown
⚙ Agent Friendliness
🔒 Security
API keys per environment. SDK keys for client-side (safe to expose). Server-side SDK key for secure flag evaluation. SOC2 Type II. Harness-owned. Feature flag data influences product behavior — access control important.
⚡ Reliability
Best When
You need feature flags tightly coupled with metrics and experimentation data — particularly when you want to measure the impact of each flag change with statistical rigor.
Avoid When
Your feature flag needs are basic, you need self-hosted control, or you're cost-sensitive — LaunchDarkly and Unleash often provide better value.
Use Cases
- • Feature flag evaluation in agent decision trees to alter behavior based on enabled features
- • Experiment configuration — creating splits and assigning traffic percentages via API
- • Reading metric impact of feature rollouts for automated reporting
- • Progressive delivery automation tied to observability metrics
- • Kill switch control for agents to disable problematic feature paths
Not For
- • Simple boolean flags without metrics (use Unleash OSS for cheaper option)
- • Teams requiring fully self-hosted feature management
- • Very small teams — pricing can be steep for limited usage
- • Non-engineering-driven feature management workflows
Interface
Authentication
Admin API keys for management operations; SDK keys (client-side and server-side) for flag evaluation. Keys are environment-scoped.
Pricing
Now integrated into Harness platform; pricing may be bundled with other Harness modules. Free tier supports basic use cases.
Agent Metadata
Known Gotchas
- ⚠ Split.io is now Harness Feature Flags — branding and documentation URLs are in transition
- ⚠ Admin API and SDK evaluation API are entirely separate surfaces with different auth
- ⚠ Treatment names are strings, not booleans — agents must compare 'on'/'off' strings, not true/false
- ⚠ Metric events for experiments require separate track() calls — not automatic from flag evaluation
- ⚠ SDK initialization is async — agents using SDK must await readiness before evaluating flags
Alternatives
Full Evaluation Report
Detailed scoring breakdown, competitive positioning, security analysis, and improvement recommendations for Split.io (Harness Feature Flags).
Scores are editorial opinions as of 2026-03-06.