komputer-ai Python SDK
Python client for the komputer.ai platform.
Installation
pip install komputer-ai-sdk
Or install directly from the repository:
pip install git+https://github.com/komputer-ai/komputer-ai.git#subdirectory=komputer-sdk/python
Quick Start
from komputer_ai.client import KomputerClient
client = KomputerClient("http://localhost:8080")
# Create an agent
agent = client.create_agent(
name="my-agent",
instructions="Summarize the latest Kubernetes release notes",
model="claude-sonnet-4-6",
)
# Stream events as the agent works
for event in client.watch_agent("my-agent"):
if event.type == "text":
print(event.payload.content)
elif event.type == "task_completed":
print(f"Done — cost: ${event.payload.cost_usd}")
break
Usage
Agents
# Create
client.create_agent(name="researcher", instructions="Research AI trends", model="claude-sonnet-4-6")
# List
agents = client.list_agents()
# Get
agent = client.get_agent("researcher")
# Update
client.patch_agent("researcher", model="claude-haiku-4-5-20251001", lifecycle="Sleep")
# Send a follow-up task to an idle agent
client.create_agent(name="researcher", instructions="Now focus on LLM benchmarks")
# Cancel a running task
client.cancel_agent_task("researcher")
# Delete
client.delete_agent("researcher")
Memories
client.create_memory(name="company-context", content="We are a B2B SaaS company.", description="Background info")
client.patch_agent("my-agent", memories=["company-context"])
memories = client.list_memories()
client.patch_memory("company-context", content="Updated context.")
client.delete_memory("company-context")
Skills
client.create_skill(name="healthcheck", description="Check service health", content="curl -s http://api/healthz")
client.patch_agent("my-agent", skills=["healthcheck"])
skills = client.list_skills()
client.delete_skill("healthcheck")
Schedules
client.create_schedule(
name="daily-report",
schedule="0 9 * * *",
instructions="Generate a daily status report",
timezone="America/New_York",
)
schedules = client.list_schedules()
client.patch_schedule("daily-report", schedule="0 10 * * *")
client.delete_schedule("daily-report")
Secrets
client.create_secret(name="api-keys", data={"GITHUB_TOKEN": "ghp_xxx", "SLACK_TOKEN": "xoxb-xxx"})
client.patch_agent("my-agent", secret_refs=["api-keys"])
secrets = client.list_secrets()
client.update_secret("api-keys", data={"GITHUB_TOKEN": "ghp_new"})
client.delete_secret("api-keys")
Connectors
client.create_connector(name="slack", service="slack", url="https://mcp.slack.com", auth_type="token")
client.patch_agent("my-agent", connectors=["slack"])
connectors = client.list_connectors()
client.update_connector("slack", token="xoxp-new-token") # rotate the token in place
client.delete_connector("slack")
Offices
offices = client.list_offices()
office = client.get_office("my-office")
Streaming Events
for event in client.watch_agent("my-agent"):
match event.type:
case "task_started":
print("Agent started working...")
case "text":
print(event.payload.content)
case "tool_call":
print(f"Using tool: {event.payload.tool}")
case "task_completed":
print(f"Done — cost: ${event.payload.cost_usd}")
break
case "error":
print(f"Error: {event.payload.error}")
break
Event types: task_started, thinking, tool_use, tool_result, text, task_completed, task_cancelled, error.
Distributed consumers — group=
By default, watch_agent opens a broadcast subscription: every connected client receives every event. If you run multiple instances of your service (e.g. 3 replicas of a Slack bot) and they all call client.watch_agent("my-agent") without further options, each instance will process every event — duplicate work.
To get queue-style delivery (each event handled by exactly one instance across your fleet), pass group=:
for event in client.watch_agent("my-agent", group="my-bot"):
...
The API uses Redis-coordinated routing to deliver each event to exactly one client per group, regardless of how many replicas connect or which API replica they hit. Pick any string for the group name (my-bot, audit-pipeline, prod-webhook-fwd).
Use broadcast for: dashboards, debugging, single-instance workers, anywhere "see everything" is the goal.
Use group= for: distributed services, webhook forwarders, anywhere you'd otherwise dedupe events yourself.
On write failure, the API retries delivery to other group members on the same replica before giving up — an event is only lost when all members on the routing replica fail simultaneously. Use client.get_agent_events(name, limit=...) to backfill on reconnect if you need strict exactly-once guarantees.
Context Manager
with KomputerClient("http://localhost:8080") as client:
agents = client.list_agents()
1 maintainer
Metadata
1 maintainer
Release files for komputer-ai-sdk 0.22.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| komputer_ai_sdk-0.22.0.tar.gz | 153.8 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| komputer_ai_sdk-0.22.0-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 534.5 kB
Release files / komputer_ai_sdk-0.22.0.tar.gz
| Download URL | komputer_ai_sdk-0.22.0.tar.gz |
|---|---|
| Size | 153.8 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
22ab2ab3722a9fc1ff5e19b454d39c7c3898042cd950a989a5133224900e51e0
|
|
BLAKE2b-256 checksum How to use checksums |
0f8f4b8f98772063c5d6a57b29523ae5d56ee8514fb37aaf45a38b3b723edd29
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Oct 13, 2026.
Transparency logRelease files / komputer_ai_sdk-0.22.0-py3-none-any.whl
| Download URL | komputer_ai_sdk-0.22.0-py3-none-any.whl |
|---|---|
| Size | 380.7 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
e5497ca27ede8b6d9f5b7210a6ddacb315a0214f869d978362dfab018ed66982
|
|
BLAKE2b-256 checksum How to use checksums |
3e725e29b0e09029d4381d21bdec2b828a6440e3ce3a432b8cd0002ee206216d
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Oct 13, 2026.
Transparency log
