MANUSH Developer Platform · MANUSH SDK
Build the skills intelligent machines will run.
With the MANUSH SDK, developers create robot behaviors, manipulation skills, navigation modules, perception modules, tools, voice agents, enterprise integrations and simulation environments — then train them in MANUSH Sim, test them against a virtual robot in the Sandbox and ship them through the MANUSH Skill Store.
Already building? Sign in to your developer dashboard.
# 1 · Create a virtual MANUSH H2 in the Sandbox
curl -X POST https://manush.si/api/v1/sandbox/robots \
-H "Authorization: Bearer $MANUSH_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "H2", "name": "line-2-humanoid"}'
# 2 · Send a SafeCore-checked command
curl -X POST https://manush.si/api/v1/sandbox/robots/SBX-H2-XXXXXXXX/commands \
-H "Authorization: Bearer $MANUSH_API_KEY" \
-H "Content-Type: application/json" \
-d '{"command": "move_to", "params": {"x": 4.0, "y": 2.5}}'
# 3 · Read simulated telemetry
curl https://manush.si/api/v1/sandbox/robots/SBX-H2-XXXXXXXX/telemetry \
-H "Authorization: Bearer $MANUSH_API_KEY"
import os, requests
API = "https://manush.si/api/v1"
auth = {"Authorization": f"Bearer {os.environ['MANUSH_API_KEY']}"}
robot = requests.post(f"{API}/sandbox/robots", headers=auth,
json={"model": "H2", "name": "line-2-humanoid"}).json()["data"]
requests.post(f"{API}/sandbox/robots/{robot['serial']}/commands", headers=auth,
json={"command": "move_to", "params": {"x": 4.0, "y": 2.5}})
tele = requests.get(f"{API}/sandbox/robots/{robot['serial']}/telemetry", headers=auth).json()["data"]
print(tele["pose"], tele["battery"]["percent"], tele["safecore"])
const API = "https://manush.si/api/v1";
const headers = { Authorization: `Bearer ${process.env.MANUSH_API_KEY}`, "Content-Type": "application/json" };
// Queue a synthetic MANUSH Sim run: H2 in a monsoon warehouse with synthetic humans
const res = await fetch(`${API}/sim/jobs`, {
method: "POST", headers,
body: JSON.stringify({
environment: "monsoon-warehouse", robot_model: "H2", episodes: 500, seed: 42,
scenario: { synthetic_humans: true, failure_injection: true, safety_testing: true },
}),
});
const { data: job } = await res.json();
console.log(job.id, job.status); // poll GET /sim/jobs/{id} for results
01 MANUSH SDK
What you can build
Developers should be able to create robot behaviors, manipulation skills, navigation modules, perception modules, tools, voice agents, enterprise integrations and simulation environments. This creates a developer ecosystem around the robots.
Robot behaviors
Compose perception, navigation and manipulation into task-level behaviours.
Manipulation skills
Grasping, tool use, assembly and handover skills for hands and arms.
Navigation modules
SLAM, mapping, path planning and site-specific navigation.
Perception modules
Detectors, segmenters and sensor-fusion pipelines.
Tools
Drivers and SafeCore-authorised end-effector tools.
Voice agents
Multilingual voice interaction across Indian languages and dialects.
Enterprise integrations
Connect robots to WMS, MES, ERP and ticketing systems.
Simulation environments
MANUSH Sim scenes, synthetic objects and scenarios.
02 MANUSH OS
Program against MANUSH OS modules
An operating system for embodied intelligence. Every skill declares which MANUSH OS modules it uses — and MANUSH Safety always sits underneath.
PERCEPTION
MANUSH Perception
Computer vision and sensor fusion.
NAVIGATION
MANUSH Navigation
SLAM, spatial mapping, planning.
MOTION
MANUSH Motion
Locomotion and whole-body control.
MANIPULATION
MANUSH Manipulation
Hands, arms and tools.
VOICE
MANUSH Voice
Natural multilingual interaction.
MEMORY
MANUSH Memory
Short-term + long-term contextual memory.
SAFETY
MANUSH Safety
Hard safety boundaries and runtime monitoring.
SKILLS
MANUSH Skills
Reusable robot capabilities.
AGENT
MANUSH Agent
High-level task reasoning.
03 The platform
Everything between an idea and a deployed skill
Developer accounts & API keys
Scoped sandbox and live keys with rate limits, rotation, revocation and per-key usage metering.
Learn more →SDK downloads
MANUSH SDK for Python, C++, ROS 2, JavaScript and the manush CLI — every build with a SHA-256 checksum.
Learn more →Documentation
Guides and references: skill manifest, SafeCore permissions, MANUSH Sim, Sandbox API and the Fleet API preview.
Learn more →Robot skills
Version skills with a validated manifest, upload signed packages and submit them for MANUSH SafeCore review.
Learn more →Simulation
Queue MANUSH Sim runs across factory, warehouse, home, city and India-first environments.
Learn more →Sandbox
Virtual robots that answer the MANUSH Cloud API shape — move, pick, place, run skills, estop — without hardware.
Learn more →Teach by demonstration
Show it once → teach the robot the task.
A person demonstrates how to perform a task. MANUSH converts the demonstration into a reusable robot skill. Describe the steps and recording metadata in the developer dashboard and MANUSH drafts the skill manifest — primitives, required sensors and a conservative SafeCore envelope — ready to train in MANUSH Sim.
How demonstration-to-skill works →- 01 Demonstrate — teleoperation, kinesthetic teaching, video or motion capture
- 02 Describe — steps, objects, tools, workspace, humans nearby
- 03 Draft — manifest with primitives, sensors and SafeCore permissions
- 04 Train & evaluate — synthetic runs in MANUSH Sim
- 05 Review & publish — SafeCore checklist, signed package, Skill Store
04 MANUSH Sim
Millions of simulated hours → fewer physical training hours.
A robotic digital-twin platform: factory simulation, warehouse simulation, home simulation, city environments, physics simulation, synthetic data, reinforcement learning, synthetic humans, synthetic objects, synthetic environments, failure simulation and safety testing.
factory
Factory simulation
Discrete-manufacturing hall: assembly cells, conveyors, fixtures, forklifts and operators working alongside robots.
warehouse
Warehouse simulation
High-bay warehouse with racking, pallets, totes, AMR traffic lanes and pick stations.
home
Home simulation
Indian apartment: kitchen, living room, bedrooms, bathroom, household objects, pets and family members.
city
City environments
Streets, sidewalks, markets and building entrances with pedestrians, two-wheelers, kerbs and mixed indoor/outdoor transi…
physics
Physics simulation
Contact-rich physics proving ground: slopes, low-friction surfaces, deformable objects, payload variation and actuator l…
synthetic
Synthetic data generation
Domain-randomised scenes (textures, lighting, camera noise, object poses) for training MANUSH Perception models with lab…
synthetic
Reinforcement learning gym
Reset-friendly task arenas with reward shaping for reinforcement learning of locomotion and manipulation policies.
synthetic
Synthetic humans
Dense populations of simulated people with varied body sizes, clothing, walking speeds and intentions — for human-aware…
05 MANUSH Skill Store
App Store for Robots
Companies purchase specialized robot skills. Developers earn a revenue share on every third-party skill.
MANUSH H2 + Automotive Assembly Skill Pack
Skill pack for MANUSH H2 on automotive lines: part kitting, fastening with torque verification, sealant application, harness routing and quality gates…
MANUSH D2 + Solar Farm Inspection Skill
MANUSH D2 Industrial patrols utility-scale PV plants: thermal detection of hotspots, cracked modules, soiling, inverter faults and perimeter breaches.
Carry — Tote & Material Carrying
Move totes, trays and materials between stations, wards and floors — including elevators — with human-aware navigation.
Clean — Facility Cleaning
Autonomous floor scrubbing, vacuuming and UV-C disinfection for hotels, hospitals, offices and homes with crowd-aware scheduling.
Inspect — Visual & Thermal Inspection
Autonomous inspection rounds with RGB, thermal and acoustic capture — hotspots, leaks, corrosion and gauge readings logged to MANUSH Cloud.
Sort — Parcel & Part Sorting
High-throughput sorting of parcels and parts by destination, SKU or defect class using vision, barcodes and weight checks.
MANUSH SafeCore™
Every skill declares its safety envelope.
High-level AI should never be allowed to bypass the low-level safety system. Skills declare SafeCore permissions — maximum speed, torque, minimum human distance and authorised tools — and MANUSH reviews each version against speed, torque, proximity, collision detection, emergency stop, restricted zones, human detection, tool authorization, battery safety, software permissioning, remote shutdown and anomaly detection before it reaches the Skill Store.
SafeCore permissions reference →Business model
Skill marketplace
Revenue share from third-party robot skills — 70 % to the developer by default.
Developer platform
SDK/API subscriptions for teams building on MANUSH.
Simulation
MANUSH Sim enterprise licensing.
06 Documentation
Start reading
MANUSH Developer Platform
The developer platform becomes the MANUSH ecosystem.
Create an account, get a sandbox API key in under a minute and ship your first skill to MANUSH robots.