# Osmosis

**Canonical:** https://apis.io/providers/osmosis/  
**Website:** https://osmosis.ai/  
**APIs profiled:** 3

Osmosis (Gulp AI Inc.) is a forward-deployed reinforcement-learning platform for post-training large language models. Teams implement an AgentWorkflow and a Grader in Python, then use the Osmosis CLI and web platform to submit evaluation and training runs (GRPO/DAPO, multi-turn tool training) that produce task-specific LoRA models which beat foundation models at a fraction of the cost. Osmosis also ships an Agent Improvement REST API for storing agent interaction knowledge and enhancing agent tasks from past interactions, an open-source Python SDK/CLI (osmosis-ai), a TypeScript logging SDK, an open-source MCP server (Osmosis-Apply), webhooks, and hosted documentation. Backed by CRV, Felicis, and Paradigm.

## Kin Score — 52.6 / 100 (developing)

Scored 2026-08-17 under rubric 0.11.0. Trend: flat (+0.0 from 52.6).

| Facet | Score |
|---|---|
| Discoverability | 81.5 |
| Contract Quality | 63.0 |
| Governance | 11.5 |
| Operational Transparency | 28.9 |
| Developer Ergonomics | 75.5 |
| Commercial Clarity | 42.1 |

## Agent readiness — 55.2 (agent-ready)

| Dimension | Value |
|---|---|
| Spec Presence | yes |
| Agentic Access | derived |
| MCP Server | yes |
| Auth Clarity | yes |
| Idempotency | no |
| Error Semantics | verified |
| OpenAPI Examples | partial |
| Rate Limit Signal | no |
| Event Surface Described | yes |
| Agent Skills | derived |
| Well Known Catalog | no |
| Consent Identity | no |
| Agent Card | no |
| Dry Run Mode | no |

## Access

Self-serve signup — onboarding: self-serve, pricing: unknown, trial: no (confidence: medium).

## APIs (3)

- **Osmosis agent API** — Operations for enhancing agent interactions and decisions
- **Osmosis knowledge API** — Operations for managing the knowledge base
- **Osmosis Osmosis Agent Improvement API API** — The Osmosis Agent Improvement API API from Osmosis — 1 operation(s) for osmosis agent improvement api.

## MCP servers (1)

- **osmosis-mcp.yml**

## Agentic access (1)

- **Osmosis Agentic Access** — 5 operations · 3 acting

## Security (2)

- **Osmosis Authentication** — apiKey · 1 scheme
- **Osmosis Domain Security** — TLSv1.3 · HSTS · DNSSEC · DMARC

## Tags

Company, AI, Reinforcement Learning, LLM, Post-Training, Model Training, Agents, Machine Learning, Developer Tools

---

Profiled by [API Evangelist](https://apievangelist.com) and published on [APIs.io](https://apis.io/providers/osmosis/). Scores are computed from the provider's own public artifacts under a published rubric.