# Agents - Obss

Set up
Agents
See every agent Obss runs, which model providers each one uses, and how to turn agents on for your experiments.
Agents, such as Claude Code or Codex, do your tasks in an experiment. You choose them at launch, so they aren’t part of a
task
(the work) or a
treatment
(the tools the agent gets). Built-in agents start turned on in each project and list models once you connect a provider they use. This is step 3 of the
setup checklist
.
Built-in agents
Agent
Vendor
Models from
Codex
OpenAI
OpenAI API key, or
OpenAI (ChatGPT Subscription)
Claude Code
Anthropic
Anthropic API key, or
Anthropic (Claude Subscription)
Cursor
Cursor
Cursor API key from a paid plan
GitHub Copilot
GitHub
GitHub
on
Profile → Integrations
Grok Build
xAI
xAI API key, or a compatible provider
Antigravity
Google
Google Gemini API key only
OpenCode
Anomaly
OpenCode Zen, OpenCode Go, or a compatible provider
Pi
Earendil Works
A compatible provider
Qwen Code
Alibaba Cloud
Alibaba Cloud Model Studio, or a compatible provider
Hermes
Nous Research
A compatible provider, or
Nous Portal
OpenClaw
OpenClaw Foundation
OpenCode Zen, OpenCode Go, or a compatible provider
Which providers each agent uses
An agent lists a model only when the model serves an API that the agent supports, so a gateway such as OpenRouter can show a model to one agent and not another. The three model APIs are OpenAI Responses, OpenAI Chat Completions, and Anthropic Messages:
Responses only:
Codex.
Messages only:
Claude Code.
Chat Completions and Messages:
Qwen Code.
All three:
GitHub Copilot, Grok Build, OpenCode, Pi, Hermes, and OpenClaw.
Their own API:
Cursor (Cursor’s) and Antigravity (Gemini).
In the table, a
compatible provider
is any organization API-key provider on
Model providers
, including OpenAI- and Anthropic-compatible ones, except Cursor, OpenCode Zen, and OpenCode Go.
Zen
and
Go
are OpenCode Zen and OpenCode Go.
Agent
Providers
Not supported
Codex
OpenAI,
OpenAI (ChatGPT Subscription)
, and Responses models on compatible providers, Zen, and Go
xAI
Claude Code
Anthropic, Anthropic-compatible,
Anthropic (Claude Subscription)
, and Messages models on compatible providers, Zen, and Go
None
Cursor
Cursor only
Every other provider
GitHub Copilot
GitHub
, compatible providers, Zen, and Go
Google Gemini API
Grok Build
xAI, compatible providers, Zen, and Go
Google Gemini API
Antigravity
Google Gemini API only
Every other provider
OpenCode
Zen, Go, and compatible providers
Google Gemini API, Cursor, personal integrations
Pi
Compatible providers, Zen, and Go
Google Gemini API, Cursor, personal integrations
Qwen Code
Alibaba Cloud Model Studio (listed first), OpenAI, Anthropic, Zen, Go, and the other direct APIs, gateways, and OpenAI-compatible providers
xAI, Google Gemini API, Anthropic-compatible, personal integrations
Hermes
Nous Portal
, compatible providers, Zen, and Go
Cursor
OpenClaw
Zen, Go, and compatible providers
Cursor, personal integrations
Some combinations have extra limits:
Google Gemini API and xAI models don’t serve Anthropic Messages, and Gemini models don’t serve Responses. So Claude Code lists no models from either, and Codex lists none from Gemini.
On OpenRouter, Claude Code uses only
anthropic/claude…
models.
Each personal integration runs only its own agent: ChatGPT Subscription runs Codex, Claude Subscription runs Claude Code, GitHub runs GitHub Copilot, and Nous Portal runs Hermes.
Grok Build can use a Claude model over Messages only when the model supports adaptive thinking, so on Anthropic it can’t use Claude Sonnet 4.5, Haiku 4.5, or Opus 4.5.
Hermes can’t use an Anthropic-compatible provider that needs custom headers.
Cursor needs a key from a paid Cursor plan, and GitHub Copilot needs a paid Copilot plan.
Turn on an agent
Agents are turned on or off per project.
Manage → Agents
lists every agent with its model providers and an
Enabled
switch.
Dashboard
CLI
Plugin (MCP)
Connect a provider that the agent uses, under
Organization settings → Model providers
or
Profile → Integrations
.
Open
Manage → Agents
.
Turn on the agent’s
Enabled
switch, then select
Save changes
.
Or open the agent and turn on
In this project
, which saves at once.
List the agents with their providers and current revision:
obss
project
ai
inspect
--view
agents
Turn agents on with the revision from that output:
obss
project
ai
apply
--operation
'{"action":"set_agents","expectedRevision":12,"agents":[{"key":"claude-code","enabled":true},{"key":"codex","enabled":true}]}'
Omit
agents
to turn every agent on or off. If the revision is stale, the command changes nothing (
obss project ai apply
).
Ask your AI assistant to turn on the agents. It calls
Get built-in and custom agents, models, and Obss AI settings
(
inspect_project_ai
), then
Create custom agents and change agents, models, and Obss AI settings
(
manage_project_ai
).
Turn on Claude Code and Codex in my Obss project.
Choose a model and effort
When you
launch an experiment
, choose an agent, one or more of its models, and an effort level if the model supports one. Levels run
none
,
minimal
,
low
,
medium
,
high
,
xhigh
,
max
, and the picker shows only the ones the model lists.
Default
sends no effort level, so the agent uses its own default. The one exception is Cursor, which uses the default variant Cursor lists for the model.
How agents run
Each run uses the agent as its vendor ships it, with the provider and model you chose. Obss changes the agent’s settings only where a run needs it:
No one answers prompts during a run, so the agent never asks for approval, asks questions, or asks you to sign in.
The agent calls only the provider and model you chose.
Agent and model versions are pinned for each run, and the agent doesn’t update itself.
The agent records more detail about its work, such as reasoning summaries.
The run ends when the agent finishes its own work. Obss waits for subagents and workflows the agent started, but not for processes it left running, such as a dev server.
With the vendor’s own API, the agent gets the features the vendor gives its direct customers. For example, Codex can search the web with an OpenAI API key on OpenAI’s endpoint or with a ChatGPT Subscription, and Grok Build can search the web only with an xAI API key.
These run settings can make runs cost more and take longer. At
Default
effort, OpenClaw applies its own reasoning level for the model, Codex and Grok Build can make billed web searches, and Claude Code waits for the subagents and workflows that it started.
An agent’s page and runs
Select an agent on
Manage → Agents
to open its page:
Tab
What it shows
Overview
A built-in agent’s
Status
,
Model providers
, and, when it’s ready,
New experiment
, which starts an experiment with that agent alone. For a custom agent, its settings.
Experiments
The experiments of the agent’s runs
Runs
Every run of the agent, with any model, in every experiment
Frictions
The
frictions
of those runs
A removed agent keeps its page and these tabs while it has runs.
From the CLI and the plugin, filter
obss experiments list
,
obss runs list
, or
obss frictions list
by the agent’s key, such as
--agent-id claude-code
(
agent_id
). A run’s
agentId
is that key. For one model, pass
--agent-model-id
(
agent_model_id
) with an
agm_
ID from
obss catalog list --kind agent
.
Custom agents
Custom agents are in private beta. Email
support@obss.app
to enable them for your organization.
A
custom agent
is an agent that your organization adds through the Agent Client Protocol (ACP) with
Connect custom agent
. Every project lists it with a
Custom
tag. It doesn’t use a model provider: its models come from the agent itself, and its credentials are its own environment variables.
From the CLI and the plugin
Task
CLI
Plugin (MCP)
List agents, their providers, and their status
obss project ai inspect --view agents
inspect_project_ai
Turn agents on or off
obss project ai apply
with
set_agents
manage_project_ai
Turn an agent’s models on or off
obss project ai apply
with
set_model_enabled
manage_project_ai
Find agent model IDs for a launch
obss catalog list --kind agent
inspect_catalog
List an agent’s runs
obss runs list --agent-id
list_runs
List an agent’s frictions
obss frictions list --agent-id
list_frictions
Next steps
Custom agents
Step 4 (optional): add your own ACP agent.
Obss AI features
Step 5: choose models for Evals, Insights, and Ask Obss.
Compare agents and models
Run one task across several agents.
Launch an experiment
Choose agents, models, and effort at launch.
Model providers
Custom agents
Powered by
This documentation is built and hosted on Mintlify, a developer documentation platform