January AI

January AI is the health intelligence infrastructure from January, Inc., which combines photo-based food recognition, device-free glucose prediction, and clinical nutrition context into a single API. Teams no longer have to build their own meal-image and glucose models, because one call returns the dish, the portion sizes, and the macronutrients.

Core capabilities

• Photo-based food recognition: a single meal photo returns the dish, the food list, and portion sizes, and the in-house vision model scores 86.24 on the open MIT benchmark
• Device-free glucose prediction: without a CGM it still returns a 2-hour curve at 15-minute intervals plus an impact level, backed by US patent US 12,749,580
• Health context engine: organizes EHR, lab, wearable, and claims data into a personal knowledge graph and longitudinal timeline across 5 coding standards
• Clinical nutrition monitoring: shows patient nutrition trends inside Epic overlaid with medication and weight, and has been selected for the Mayo Clinic Platform and the CMS Medicare App Library
• Developer delivery: an OpenAI Chat Completions subset (the model field only accepts january), 4 SDKs, an MCP server, and short-lived client tokens

Use cases Digital health teams working on GLP-1 and weight management, prediabetes intervention products, clinical nutrition, and hospital systems. A front-end team can integrate it directly, with no need to hire a separate nutrition data team.

Unique advantages Most nutrition APIs stop at the food nutrition table, while January AI links nutrition, metabolic prediction, and clinical context into one chain. But the claim the site keeps citing, that 74.4% of participants lost weight, is only the share of the subset with at least two weight records in the paper, where actual average weight loss was about 3.3 lbs, so any citation must fall back to the original paper's framing.

Editorial assessment It is one of the few health APIs whose results can be re-run and verified with your own data, and the engineering completeness of its meal photos and glucose prediction is higher than that of comparable products. It suits US teams that want to add metabolic capabilities quickly while riding the Epic and Mayo ecosystems; anyone handling PHI must upgrade to Enterprise, because Free and Pro are not HIPAA-compliant. Overages return 429 directly with no Retry-After header, and the site has no public status page either, so load-test before signing at scale.