agents_create
ChatGPTUse to create one managed agent in Brainbase. Supply a stable idempotency key; omit runtime settings to use server defaults.
Developer Tools
Overview
Brainbase MCP lets users create, inspect, update, test, deploy, and monitor managed AI agents from ChatGPT and Codex. It supports revision-safe agent changes, registry skills and MCP server configuration, evaluations, orchestrations, schedules, and task runs, with explicit confirmation for destructive actions and OAuth-based access to the user's Brainbase workspace.
Tools
agents_delete
ChatGPTUse only after explicit user confirmation. Permanently deletes the agent and requires its latest revision and exact title.
agents_get
ChatGPTUse to read an agent's current configuration and opaque revision immediately before a targeted update or destructive action.
agents_get_revision
ChatGPTUse to fetch only the agent's latest opaque revision when its full configuration was already inspected and remains current.
agents_list
ChatGPTUse to find managed agents the user can access, optionally within one resolved team, before inspecting or modifying an agent.
agents_update
ChatGPTUse to overwrite selected agent profile fields with the latest expected revision. Existing tasks keep their runtime snapshot.
evals_create
ChatGPTUse to add one evaluation definition to an agent. The definition does not run until evals_run is called.
evals_delete
ChatGPTUse only after explicit user confirmation. Permanently deletes one eval definition and requires its latest revision and exact slug.
evals_get
ChatGPTUse to read one evaluation definition and its opaque revision before updating, deleting, or running it.
evals_list
ChatGPTUse to discover evaluation definitions for one agent before inspecting, updating, running, or deleting an eval.
evals_results
ChatGPTUse to inspect stored evaluation verdicts for an agent, optionally limited to one task created by an eval run.
evals_run
ChatGPTUse to start a billable agent task for one enabled evaluation. Disclose billable execution before calling and use an idempotency key.
evals_update
ChatGPTUse to overwrite selected fields on an evaluation definition with its latest opaque revision.
instructions_update
ChatGPTUse when only an agent's instructions should change. Requires the fresh opaque revision and replaces the full instruction text.
mcp_servers_list
ChatGPTUse to inspect an agent's configured MCP servers. Credential values are never returned; only their presence is indicated.
mcp_servers_remove
ChatGPTUse to remove one named MCP server from an agent after inspecting the current configuration and obtaining a fresh revision.
mcp_servers_upsert
ChatGPTUse to add or overwrite one named, non-secret MCP server configuration while preserving hidden stored credentials.
orchestration_edges_remove
ChatGPTUse to remove one exact directed edge. Run dry_run first and supply the latest orchestration revision.
orchestration_edges_upsert
ChatGPTUse to add or overwrite one directed agent-to-agent edge. Run dry_run first and supply the latest orchestration revision.
orchestration_members_add
ChatGPTUse to add one existing agent to an orchestration. Run with dry_run first when the surrounding graph state needs validation.
orchestration_members_remove
ChatGPTUse to remove one agent and dependent graph links. Run dry_run first to review the edges and schedule links that would also be removed.
orchestrations_create
ChatGPTUse to create one private orchestration in a resolved team with a stable idempotency key. It does not run agents by itself.
orchestrations_delete
ChatGPTUse only after explicit user confirmation. Permanently deletes the orchestration and requires its latest revision and exact name.
orchestrations_get
ChatGPTUse to read an orchestration's members, edges, schedules, and fresh opaque revision before any graph change.
orchestrations_list
ChatGPTUse to discover orchestrations in one resolved team before inspecting or changing an orchestration graph.
orchestrations_update
ChatGPTUse to overwrite selected orchestration metadata with a fresh revision without replacing members, edges, or schedules.
orgs_list
ChatGPTUse when the user needs to discover the Brainbase organizations they can access before selecting a team or resource.
playbooks_archive
ChatGPTUse only after explicit confirmation to remove one active playbook. Requires the current revision and exact playbook title.
playbooks_list
ChatGPTUse to inspect an agent's active playbooks and current manifest revision before changing one playbook.
playbooks_upsert
ChatGPTUse to add or overwrite one playbook while preserving unrelated manifest components. Requires revision and idempotency guards.
schedules_remove
ChatGPTUse to remove one schedule from an orchestration after inspecting the graph and obtaining its latest revision.
schedules_test
ChatGPTUse to manually test one persisted schedule. This can start billable agent runs, so disclose that before calling.
schedules_upsert
ChatGPTUse to add or overwrite one orchestration schedule. New schedules default to inactive unless the user explicitly requests activation.
skills_attach
ChatGPTUse to attach or replace one registry skill on an agent after validating the package and obtaining the latest agent revision.
skills_detach
ChatGPTUse to remove one exact registry skill attachment from an agent after inspecting the agent and its latest revision.
skills_get
ChatGPTUse after skill search to inspect one exact creator and slug before attaching it to an agent.
skills_search
ChatGPTUse to search Brainbase's published skill registry before choosing a skill package to inspect or attach.
tasks_create
ChatGPTUse to create and run a task with one task-specific user prompt. The billable run starts by default and may invoke the agent's external tools, so disclose that before calling; set auto_run false only when the user wants a draft.
tasks_events
ChatGPTUse to inspect task activity as event types plus sanitized JSON details. Internal IDs, credentials, raw payloads, and reasoning are removed.
tasks_followup
ChatGPTUse to append one task-specific user message. Setting run starts a billable turn that may invoke the agent's external tools.
tasks_get
ChatGPTUse to read the selected task's user-facing identity and current status without returning sandbox or account diagnostics.
tasks_interrupt
ChatGPTUse only when the user wants to stop active work on a task. Interrupting can prevent the current run from completing.
tasks_list
ChatGPTUse to discover the user's Brainbase tasks, optionally for one agent, before inspecting status or events.
teams_list
ChatGPTUse after resolving an organization to list its Brainbase teams and obtain the exact team ID for agent or orchestration work.
templates_get
ChatGPTUse after template search to resolve one exact published template, its version, defaults, components, and compatibility warnings.
templates_search
ChatGPTUse to search Brainbase's published agent-template registry before creating an agent from a specific template.
Capabilities
Writes
Example Prompts
Click any prompt to copy it.
Links
App Stats
46
Tools
3
Prompts
ChatGPT
Platforms
Category
Task AutomationWorks with
ChatGPT
Data refreshed daily