---
title: "Guild vs Cursor Agent SDK"
description: "Guild vs the Cursor Agent SDK: agent projects, tools, credentials, channels, schedules, evals, and pricing compared, plus how to move an agent to Guild."
canonical: "https://www.guild.ai/compare/cursor"
---

# Guild vs Cursor Agent SDK

Guild vs the Cursor Agent SDK: agent projects, tools, credentials, channels, schedules, evals, and pricing compared, plus how to move an agent to Guild.

Cursor's Agent SDK defines an agent as a TypeScript and Markdown project you deploy to Cursor or serve yourself. Guild runs agents in five formats, keeps every credential behind a proxy with rules set by admins, and shares agents on a public Agent Hub.

Last verified 2026-11-06. Across 35 features, Guild leads on 13, Cursor Agent SDK on 5, and 17 are level. Every cell below names its source.

## When to choose Cursor Agent SDK

- Your team already uses Cursor and wants agents billed through the same plan.
- You want each agent served with its own HTTP API and MCP endpoint.
- You want to run agents on your own Node host with agent-sdk serve.
- You want Slack, GitHub, GitLab, Bitbucket, or PagerDuty channels built into the project.

## When to choose Guild

- You want agents in more than TypeScript: instructions only, Goose, repo agents, or Python LangGraph.
- Agents must never read secrets from the environment; credentials stay behind a proxy.
- Admins, not agent code, should decide which actions each agent can take.
- You want to share agents and skills publicly on an Agent Hub others can install or fork.

## What Cursor Agent SDK and Guild are

**Cursor Agent SDK.** The Agent SDK is the harness behind Cursor's Bugbot and Security Review, opened up for your own agents. A project holds instructions, typed tools, skills, MCP connections, sub-agents, Slack and Git channels, cron schedules, and evals. Every served agent gets an HTTP API and an MCP endpoint, and models are routed and billed through Cursor.

**Guild.** Guild runs agents written as instructions, TypeScript, Goose recipes, repo agents, or Python LangGraph. Agents never see credentials: integrations go through a proxy, and admins set ALLOW/DENY rules on each operation. Agent code runs with no direct network access, and teams share agents on the Agent Hub.

|   | Cursor Agent SDK | Guild |
| --- | --- | --- |
| Category | Agents-as-code framework with managed or self-hosted serving | One platform to run and control AI agents |
| License | Proprietary (no official open-source repo found) | Proprietary platform; CLI published on npm as @guildai/cli |
| Pricing | Billed through Cursor plans (Free to Ultra $200/month, Teams $30+/user, Enterprise custom); hosting has no separate published price | No public price list; plans priced with the Guild team; model usage metered at list price |
| Maturity | Generally available (2026-10-20) | Generally available |
| Docs | https://cursor.com/docs | https://docs.guild.ai |

## Concept map: Cursor Agent SDK to Guild

| Concept | Cursor Agent SDK | Guild | Note |
| --- | --- | --- | --- |
| The agent | bot/agent.ts (defineAgent({ model, tools })) | agent.ts (agent({ description, inputSchema, outputSchema, tools, run })) | Guild adds typed agent input and output on top of typed tools. |
| Instructions | bot/instructions.md | The agent's description and prompt, or PROMPT.md in a Native agent |   |
| Tools | bot/tools/<name>.ts (defineTool with a Zod inputSchema) | Tools in a TypeScript agent, with zod schemas |   |
| Connections | bot/mcp-connections/ and host-connections/ (defineConnection) | Guild integrations (guildai~github, guildai~slack, …) or a custom integration from an MCP server or OpenAPI spec |   |
| Tool rules | tools / disallowedTools on a connection | Credential policies (ALLOW/DENY by operation), set by admins outside the agent's code |   |
| Secrets | agent-sdk secret set and ${NAME} header templates | Credentials behind the credential proxy; workspace variables via task.env |   |
| Sub-agents | bot/subagents/ and agent connections | Sub-agents called with guildAgentTool |   |
| Skills | bot/skills/ (SKILL.md) | Guild skills on the Hub (guild skill create), public or private |   |
| Schedules | bot/schedules/*.md (cron: frontmatter) | guild trigger create --type time --frequency CRON |   |
| Channels | Channels (slackChannel(), githubChannel(), defineChannel) | Slack and GitHub event triggers, API triggers | Keep channels as adapters; the agent stays the same. |
| Approvals | needsApproval on a tool | Approval gates set in the Guild UI, per agent and per action |   |
| Evals | evals/**/*.eval.ts (defineEval, t.judge) | guild agent test and Guild Evaluations (hasToolCall, toolCallContains, llmJudge) |   |
| Deploying | agent-sdk deploy | guild agent save --wait --publish, then guild workspace agent add |   |
| Logs and traces | Sessions, agent-sdk logs, OpenTelemetry | Sessions and the session event log (guild session, guild agent logs) |   |
| Roles | Viewer / Editor / Owner roles on a deployment | Workspaces and roles |   |

## Feature by feature: Cursor Agent SDK vs. Guild

### Authoring

| Feature | Cursor Agent SDK | Guild |
| --- | --- | --- |
| Code-first TypeScript agents | **Yes.** Agents are TypeScript plus Markdown projects: `bot/agent.ts` calls `defineAgent`, and tools, skills, connections, channels and schedules come from `@grok/agent-sdk/*` helpers. ([source](https://cursor.com/docs/agent-sdk/reference/project-layout), checked 2026-11-06) | **Yes.** @guildai/agents-sdk: agent(), llmAgent(), and the task API (task.llm, task.tools, task.ui, task.gather). ([source](https://docs.guild.ai/packages/agents-sdk.md), checked 2026-10-23) |
| Python / LangGraph agents | **No.** You can't author Agent SDK agents in Python or LangGraph. A Python client (`pip install cursor-sdk`) can call Cursor's own agent, but it can't define custom agent logic. ([source](https://cursor.com/docs/sdk/python), checked 2026-11-06) | **Yes.** LangGraph agent type: langgraph.json + graph.py using guild_model() and guild_tools(). LangChain v1, Python 3.13, no extra pip packages. ([source](https://docs.guild.ai/guide/agent-types.md), checked 2026-10-23) |
| Instruction-only (no-code) agents | **Yes.** This sits outside the SDK. Cursor Automations take a prompt, model and trigger in the dashboard, and Grok Bot takes instructions, skills and tools in an app. The Agent SDK docs position both as the no-code alternatives. ([source](https://cursor.com/docs/agent-sdk), checked 2026-11-06) | **Yes.** Native agents: PROMPT.md + guild.yaml, no container, ready immediately. ([source](https://docs.guild.ai/guide/agent-types.md), checked 2026-10-23) |
| Several agent formats on one platform | **Partial.** Cursor offers Automations, Grok Bot, Agent SDK projects and Cursor SDK scripts, and `cursorPlugin()` can mount a vendored Cursor or Claude plugin. Third-party frameworks such as LangGraph or CrewAI can't be run. ([source](https://cursor.com/docs/agent-sdk/reference/extensions), checked 2026-11-06) | **Yes.** Native, TypeScript, Goose, OpenClaw, and LangGraph agent types, plus ADK (CLI 0.28). ([source](https://docs.guild.ai/guide/agent-types.md), checked 2026-10-23) |
| Typed agent input/output | **Partial.** Tool `inputSchema` (Zod) is validated, and channel routes declare Zod `bodySchema`/`querySchema` that the host validates. Agent replies are text with no declared output schema, and a tool's `outputSchema` is advertised but not validated. ([source](https://cursor.com/docs/agent-sdk/reference/channels), checked 2026-11-06) | **Yes.** Zod inputSchema and outputSchema on agent(); root input must be z.object. ([source](https://docs.guild.ai/packages/agents-sdk.md), checked 2026-10-23) |
| Repo / workspace agents (shell, git, files) | **Yes.** Cloud agents run in isolated VMs with the repo cloned, shell and git, and they push branches and PRs. Agent SDK agents get per-session workspaces with the `shell` built-in tool and can launch cloud agents via `cursorCloudAgents()`. ([source](https://cursor.com/docs/agent-sdk/guides/cloud-agents), checked 2026-11-06) | **Yes.** OpenClaw agents (AGENTS.md + skills/) run in the guildai~lobsterpot container with shell, git, and file editing. ([source](https://docs.guild.ai/guide/agent-types.md), checked 2026-10-23) |
| Sub-agents and agent-to-agent calls | **Yes.** `bot/subagents/<id>/` defines child agents with their own instructions and model. Peer agents connect via `defineConnection({ agent: "<slug>" })`, which exposes `ask`, `check` and `call_tool`. ([source](https://cursor.com/docs/agent-sdk/guides/agent-to-agent), checked 2026-11-06) | **Yes.** guildAgentTool calls owner~agent; Native, Goose, and OpenClaw agents declare sub_agents in guild.yaml. ([source](https://docs.guild.ai/guide/guild-yaml.md), checked 2026-10-23) |

### Tools & integrations

| Feature | Cursor Agent SDK | Guild |
| --- | --- | --- |
| Managed integrations catalog | **Partial.** There are built-in host APIs (`ctx.host.github`, `ctx.host.slack`), prebuilt channels (Slack, GitHub, GitLab, Bitbucket, PagerDuty) and `cursorAccount: true` connections that reuse MCP connectors authorized in the Cursor dashboard. The Marketplace lists plugins (MCP plus skills). There is no broad catalog of managed API connectors. ([source](https://cursor.com/docs/agent-sdk/reference/connections), checked 2026-11-06) | **Yes.** 37 non-model integrations (GitHub, Slack, Gmail, Linear, Jira, Notion, Zendesk, …) plus 8 model publishers. ([source](https://docs.guild.ai/integrations/overview.md), checked 2026-10-23) |
| Use MCP servers as tools | **Yes.** `bot/mcp-connections/<name>.ts` declares HTTP/SSE or stdio MCP servers via `defineConnection`. `advertiseTools: true` exposes their tools to the model. ([source](https://cursor.com/docs/agent-sdk/reference/connections), checked 2026-11-06) | **Yes.** Build custom integrations from MCP servers with guild integration. ([source](https://docs.guild.ai/integrations/overview.md), checked 2026-10-23) |
| Platform exposed over MCP | **Yes.** Each served agent exposes streamable-HTTP MCP at `/v1/mcp` (`ask`/`check`/`call_tool`) and `/v1/mcp/tools`. `agent-sdk mcp install --url` adds an agent to `~/.cursor/mcp.json`. ([source](https://cursor.com/docs/agent-sdk/reference/http-api), checked 2026-11-06) | **Yes.** Guild is an MCP server: https://api.guild.ai/mcp (OAuth + PKCE) or local `guild mcp`. ([source](https://docs.guild.ai/api-reference/mcp.md), checked 2026-10-23) |
| Custom APIs from OpenAPI | **No.** No OpenAPI-to-tools import is documented. Tools are hand-written `defineTool` files or come from MCP servers. ([source](https://cursor.com/docs/agent-sdk/reference/tools), checked 2026-11-06) | **Yes.** guild integration builds tools from OpenAPI specs; Porter drafts integrations from public APIs. ([source](https://docs.guild.ai/integrations/overview.md), checked 2026-10-23) |
| Secrets never reach the agent | **Partial.** `${NAME}` header templates and `auth` callbacks resolve on the host, so the model never sees secrets. `cursorAccount` connectors run on Cursor's backend, so tokens never reach the host. Server tools run in-process with full `process.env` access, though, so agent code can read secrets. ([source](https://cursor.com/docs/agent-sdk/reference/connections), checked 2026-11-06) | **Yes.** A credential proxy injects auth server-side; GitHub uses short-lived GitHub App tokens. ([source](https://docs.guild.ai/platform/credentials.md), checked 2026-10-23) |
| Per-operation credential policies | **Partial.** Each connection has a code-level `tools` allowlist, a `disallowedTools` denylist (deny wins) and `effects: "read"`. `host-connections/` keeps a credential invisible to the model behind narrow tools. This is set in agent code, not as an admin policy on the credential. ([source](https://cursor.com/docs/agent-sdk/reference/connections), checked 2026-11-06) | **Yes.** ALLOW/DENY rules by operation, resource, agent, and workspace; DENY wins. ([source](https://docs.guild.ai/platform/credential-policies.md), checked 2026-10-23) |
| Act as the invoking user | **Partial.** A per-session `auth` callback can return headers per caller ("gives each caller its own downstream identity"), and `ctx.session.auth` and `ctx.approvedBy` carry caller identity. Host OAuth and `cursorAccount` connections are shared by everyone who can use the agent, and no built-in per-user OAuth consent flow is documented. ([source](https://cursor.com/docs/agent-sdk/guides/mcp-oauth), checked 2026-11-06) | **Yes.** Member-scoped credentials plus identity mapping for Slack and GitHub users. ([source](https://docs.guild.ai/platform/credentials.md), checked 2026-10-23) |

### Runtime & triggers

| Feature | Cursor Agent SDK | Guild |
| --- | --- | --- |
| Managed hosted runtime | **Yes.** `agent-sdk deploy` builds the pushed git ref, runs it on Cursor's infrastructure and prints a stable URL. Deployments are managed in the Deployed Agents dashboard. ([source](https://cursor.com/docs/agent-sdk/deployment), checked 2026-11-06) | **Yes.** Guild hosts and runs all agent types. ([source](https://docs.guild.ai/platform/security-architecture.md), checked 2026-10-23) |
| Self-host option | **Yes.** `agent-sdk serve --dir ... --state-root ... --bearer-token ...` runs agents on your own Node 22.13+ host. No official container image is documented. Cloud agents can also use Self-Hosted Machines and worker pools. ([source](https://cursor.com/docs/agent-sdk/deployment), checked 2026-11-06) | **No.** Hosted only; no self-host option documented. ([source](https://docs.guild.ai/platform/security-architecture.md), checked 2026-10-23) |
| Sandboxed, network-controlled execution | **Yes.** Cloud agents run in isolated AWS VMs with egress modes ('Allow all', 'Default + allowlist', 'Allowlist only') that Enterprise can lock. Hosted Agent SDK deployments declare `hosting.egressDomains` within the team egress mode. The local sandbox is off by default. ([source](https://cursor.com/docs/cloud-agent/security-network), checked 2026-11-06) | **Yes.** Containers lose network access before the agent runs; only Guild proxies can reach out. ([source](https://docs.guild.ai/platform/security-architecture.md), checked 2026-10-23) |
| Durable long-running runs | **Partial.** Sessions, `ctx.host.kv` and `ctx.host.files` survive restarts and redeploys, and opt-in `reminders_create` tools allow later wake-ups. No pause/resume of an in-flight turn is documented, and hosted schedule handlers must finish within 26 minutes. ([source](https://cursor.com/docs/agent-sdk/reference/sessions), checked 2026-11-06) | **Partial.** Multi-turn sessions; they can be interrupted but not resumed. ([source](https://docs.guild.ai/llms.txt), checked 2026-10-23) |
| Schedules (cron) | **Yes.** `bot/schedules/` holds Markdown files with `cron:` frontmatter or `defineSchedule` handlers. Cron has 5 fields and runs in UTC, with up to 64 schedules per deploy. ([source](https://cursor.com/docs/agent-sdk/reference/schedules), checked 2026-11-06) | **Partial.** CLI supports --frequency CRON --cron-expression; docs list hourly–monthly presets. ([source](https://docs.guild.ai/platform/triggers.md), checked 2026-10-23) |
| Webhook, event & API triggers | **Yes.** There are GitHub, GitLab, Bitbucket and PagerDuty event channels, custom `defineChannel` HTTP routes and webhooks from any provider. The HTTP session API (`POST /v1/session`) is also available. Hosted routes return `202` immediately. ([source](https://cursor.com/docs/agent-sdk/reference/channels), checked 2026-11-06) | **Yes.** Event triggers (Slack mention, GitHub PR, Zendesk ticket, …) and API triggers. ([source](https://docs.guild.ai/platform/triggers.md), checked 2026-10-23) |
| Chat channels (Slack, Teams, …) | **Yes.** `slackChannel()` is prebuilt (mentions, DMs, channel posts, streaming replies), and `agent-sdk slack create` provisions the app. Teams is not documented, and other platforms such as Discord need custom channels. ([source](https://cursor.com/docs/agent-sdk/guides/slack), checked 2026-11-06) | **Yes.** Slack event triggers; Discord, Telegram, and Twilio integrations. ([source](https://docs.guild.ai/integrations/overview.md), checked 2026-10-23) |

### Control & governance

| Feature | Cursor Agent SDK | Guild |
| --- | --- | --- |
| Per-tool-call approval gate | **Partial.** The tools reference mentions `needsApproval` calls, with `ctx.approvedBy` and a Slack Approve button, but none of the pages I read explain how to configure it. Hooks are observe-only, and headless Cursor SDK local runs execute tools without approval. ([source](https://cursor.com/docs/agent-sdk/reference/tools), checked 2026-11-06) | **Yes.** Approval gates are set in the Guild UI, per agent and per action: when the agent reaches a gated action, the run pauses until a designated human approves. No code needed. ([source](https://www.guild.ai/product/ai-governance), checked 2026-10-23) |
| Roles & access control | **Yes.** Each deployed agent has Viewer, Editor (`deployer`) and Owner (`admin`) roles. Team Admins can manage every bot the team owns, and only Owners change access. ([source](https://cursor.com/docs/agent-sdk/reference/access), checked 2026-11-06) | **Partial.** Admin and Member org roles, restricted workspaces, and scoped API keys. ([source](https://docs.guild.ai/llms.txt), checked 2026-10-23) |
| Audit log | **Partial.** Enterprise audit logs cover admin and security events (sign-ins, roles, API keys, settings, integrations) via the dashboard, `GET /teams/audit-logs` or SIEM streaming. They exclude agent prompts and responses, and the docs don't say whether Agent SDK deploys are logged. ([source](https://cursor.com/docs/enterprise/compliance-and-monitoring), checked 2026-11-06) | **Yes.** Immutable, CSV-exportable audit log (admins). ([source](https://docs.guild.ai/llms.txt), checked 2026-10-23) |
| Runtime guardrails & scanning | **Partial.** No PII or content-scanning guardrail API is documented. Cursor Agent hooks (`.cursor/hooks.json`, e.g. `beforeShellExecution`, `preToolUse`) can block or modify the agent loop, and an Auto-review classifier (best-effort) exists for the Cursor SDK. Agent SDK `defineHook` is observe-only. ([source](https://cursor.com/docs/agent-sdk/reference/hooks), checked 2026-11-06) | **No.** No prompt or tool-call scanning documented; isolation comes from the network-less runtime and credential policies. ([source](https://docs.guild.ai/platform/security-architecture.md), checked 2026-10-23) |
| Spend budgets & usage limits | **Partial.** Enforced spend caps exist only at account or team level: a monthly additional-usage limit, `team_hard_limit_dollars` and per-user limits. No per-agent token or cost budget is documented. `concurrency.maxRunningTurns` caps concurrent turns. ([source](https://cursor.com/docs/models-and-pricing), checked 2026-11-06) | **Yes.** Per-agent and per-workspace spend budgets, model allowlists, and daily token limits. ([source](https://docs.guild.ai/llms.txt), checked 2026-10-23) |
| Multi-provider models with fallbacks | **Partial.** Models from OpenAI, Anthropic, Google, xAI (Grok, the default `grok-4.5`) and Cursor Composer are available, set via `defineAgent({ model })`. Auto and Cursor Router pick models server-side, but no configurable fallback chain is documented, and all inference runs through Cursor (no BYO endpoint). ([source](https://cursor.com/docs/agent-sdk/reference/agent-config), checked 2026-11-06) | **Yes.** guild.yaml models is an ordered fallback list across Anthropic, OpenAI, Gemini, Meta, DeepSeek, and more; BYO keys supported. ([source](https://docs.guild.ai/guide/guild-yaml.md), checked 2026-10-23) |

### Observability

| Feature | Cursor Agent SDK | Guild |
| --- | --- | --- |
| Run traces | **Yes.** Sessions keep an append-only record of messages and tool calls, viewable via `agent-sdk session get --view trajectory`, `agent-sdk logs` and the playground `traceUrl`. OpenTelemetry export of traces, metrics and logs uses `OTEL_EXPORTER_OTLP_ENDPOINT`. ([source](https://cursor.com/docs/agent-sdk/guides/opentelemetry), checked 2026-11-06) | **Yes.** Session event log of LLM calls, tool calls, sub-tasks, and container logs; also via the API. ([source](https://docs.guild.ai/llms.txt), checked 2026-10-23) |
| Evaluations | **Yes.** `evals/**/*.eval.ts` files use `defineEval` with tool-call assertions, Zod shape `matches`, and LLM judge (`t.judge.factuality`). `agent-sdk eval` runs locally or in CI with JUnit output, and `github replay` saves fixtures. ([source](https://cursor.com/docs/agent-sdk/evals), checked 2026-11-06) | **Partial.** Evaluations with samples, checks, and llmJudge; enabled per account, no CLI command yet. ([source](https://docs.guild.ai/platform/evaluations.md), checked 2026-10-23) |
| Cost & usage tracking | **Yes.** `GET /v1/session/:id/cost` and `agent-sdk session list --view cost` report per-session cost. The Cursor SDK `getUsage()` returns charged cents, and the `cursor.token.usage` OTel metric is emitted. ([source](https://cursor.com/docs/agent-sdk/reference/http-api), checked 2026-11-06) | **Yes.** Insights and Spend Explorer. ([source](https://docs.guild.ai/llms.txt), checked 2026-10-23) |

### Ecosystem & business

| Feature | Cursor Agent SDK | Guild |
| --- | --- | --- |
| Public agent hub (install & fork) | **Partial.** The Cursor Marketplace lists plugins (MCP servers plus skills) and prebuilt Automations, and the Origin App Marketplace lists reviewed third-party apps. There is no public hub for installing and forking Agent SDK agents (the access docs only mention 'catalog' bots in passing). Extensions support overlays but not forks. ([source](https://cursor.com/marketplace), checked 2026-11-06) | **Yes.** Agent Hub: Draft, Internal, or Public; owner~name with semver; install and fork (guild agent init --fork). ([source](https://docs.guild.ai/platform/publish-to-agent-hub.md), checked 2026-10-23) |
| Reusable skills | **Yes.** `bot/skills/` takes flat `.md`, `<name>/SKILL.md` directories or `defineSkill` TS. Skills are advertised by description and loaded on demand. ([source](https://cursor.com/docs/agent-sdk/reference/skills), checked 2026-11-06) | **Yes.** Hub skills, `guild skill`, and skills inside Goose and OpenClaw agents. ([source](https://docs.guild.ai/cli/commands.md), checked 2026-10-23) |
| Full lifecycle CLI | **Yes.** `agent-sdk` covers init, dev, run, chat, eval, validate, deploy, the deployment start/stop/rollback/upgrade family, secret, logs, session, mcp, slack and github replay. ([source](https://cursor.com/docs/agent-sdk/reference/cli), checked 2026-11-06) | **Yes.** 17 top-level commands: init, test, save --publish, workspace chat, trigger, integration, skill, session, and more. ([source](https://docs.guild.ai/cli/commands.md), checked 2026-10-23) |
| Open-source core | **No.** No official open-source repository for the Agent SDK was found, and third-party reports describe Cursor's SDK packages as proprietary under the Cursor Terms. I couldn't open the npm page (403) to confirm the license. ([source](https://dev.to/alanwest/cursor-agent-sdk-vs-claude-agent-sdk-every-model-is-billed-through-cursor-gal), checked 2026-11-06) | **No.** Proprietary platform; the CLI is published on npm. ([source](https://docs.guild.ai/llms.txt), checked 2026-10-23) |
| Public pricing | **Partial.** Plan and per-token model prices are public. Pricing for Agent SDK hosting and deployed agents is not itemized anywhere I found; usage appears to draw from plan usage and credits. ([source](https://cursor.com/pricing), checked 2026-11-06) | **Partial.** No public price list: plans are priced with the Guild team; model usage is metered at list price. ([source](https://docs.guild.ai/reference/pricing.md), checked 2026-10-23) |

## Frequently asked questions

### Is Cursor's Agent SDK the same as the Cursor SDK?

No. The Cursor SDK (public beta since April 2026) is a client library that drives Cursor's own coding agent, locally or in a cloud VM. The Agent SDK (generally available since October 2026) is for defining and deploying your own agents.

### How do I move a Cursor Agent SDK agent to Guild?

Create a TypeScript agent with guild agent init --agent-type GUILD\_TYPESCRIPT. Move instructions.md into the agent's description, each tools/ file into a typed tool, and MCP connections into Guild integrations. Set secrets as Guild credentials, recreate schedules with guild trigger create, and turn evals into guild agent test cases or Guild Evaluations. Use the concept map above as the checklist.

### Can I self-host?

Cursor's agents can run on your own Node host with agent-sdk serve. Guild is hosted only.

### What doesn't map 1:1?

Each Cursor agent's own HTTP and MCP endpoints, the Slack and Git channels built into the project, and shared journal memory. On Guild, agents are reached through workspace chat, triggers, and Guild's MCP server, and channels are integrations.

### Is the Agent SDK's pricing public?

Partly. Cursor's plans and model prices are public, but hosting for deployed agents has no separate published price.

## Sources

- [cursor.com/docs/agent-sdk/reference/project-layout](https://cursor.com/docs/agent-sdk/reference/project-layout)
- [cursor.com/docs/sdk/python](https://cursor.com/docs/sdk/python)
- [cursor.com/docs/agent-sdk](https://cursor.com/docs/agent-sdk)
- [cursor.com/docs/agent-sdk/reference/extensions](https://cursor.com/docs/agent-sdk/reference/extensions)
- [cursor.com/docs/agent-sdk/reference/channels](https://cursor.com/docs/agent-sdk/reference/channels)
- [cursor.com/docs/agent-sdk/guides/cloud-agents](https://cursor.com/docs/agent-sdk/guides/cloud-agents)
- [cursor.com/docs/agent-sdk/guides/agent-to-agent](https://cursor.com/docs/agent-sdk/guides/agent-to-agent)
- [cursor.com/docs/agent-sdk/reference/connections](https://cursor.com/docs/agent-sdk/reference/connections)
- [cursor.com/docs/agent-sdk/reference/http-api](https://cursor.com/docs/agent-sdk/reference/http-api)
- [cursor.com/docs/agent-sdk/reference/tools](https://cursor.com/docs/agent-sdk/reference/tools)
- [cursor.com/docs/agent-sdk/guides/mcp-oauth](https://cursor.com/docs/agent-sdk/guides/mcp-oauth)
- [cursor.com/docs/agent-sdk/deployment](https://cursor.com/docs/agent-sdk/deployment)
- [cursor.com/docs/cloud-agent/security-network](https://cursor.com/docs/cloud-agent/security-network)
- [cursor.com/docs/agent-sdk/reference/sessions](https://cursor.com/docs/agent-sdk/reference/sessions)
- [cursor.com/docs/agent-sdk/reference/schedules](https://cursor.com/docs/agent-sdk/reference/schedules)
- [cursor.com/docs/agent-sdk/guides/slack](https://cursor.com/docs/agent-sdk/guides/slack)
- [cursor.com/docs/agent-sdk/reference/access](https://cursor.com/docs/agent-sdk/reference/access)
- [cursor.com/docs/enterprise/compliance-and-monitoring](https://cursor.com/docs/enterprise/compliance-and-monitoring)
- [cursor.com/docs/agent-sdk/reference/hooks](https://cursor.com/docs/agent-sdk/reference/hooks)
- [cursor.com/docs/models-and-pricing](https://cursor.com/docs/models-and-pricing)
- [cursor.com/docs/agent-sdk/reference/agent-config](https://cursor.com/docs/agent-sdk/reference/agent-config)
- [cursor.com/docs/agent-sdk/guides/opentelemetry](https://cursor.com/docs/agent-sdk/guides/opentelemetry)
- [cursor.com/docs/agent-sdk/evals](https://cursor.com/docs/agent-sdk/evals)
- [cursor.com/marketplace](https://cursor.com/marketplace)
- [cursor.com/docs/agent-sdk/reference/skills](https://cursor.com/docs/agent-sdk/reference/skills)
- [cursor.com/docs/agent-sdk/reference/cli](https://cursor.com/docs/agent-sdk/reference/cli)
- [dev.to/alanwest/cursor-agent-sdk-vs-claude-agent-sdk-every-model-is-billed-through-cursor-gal](https://dev.to/alanwest/cursor-agent-sdk-vs-claude-agent-sdk-every-model-is-billed-through-cursor-gal)
- [cursor.com/pricing](https://cursor.com/pricing)
- [docs.guild.ai/packages/agents-sdk](https://docs.guild.ai/packages/agents-sdk.md)
- [docs.guild.ai/guide/agent-types](https://docs.guild.ai/guide/agent-types.md)
- [docs.guild.ai/guide/guild-yaml](https://docs.guild.ai/guide/guild-yaml.md)
- [docs.guild.ai/integrations/overview](https://docs.guild.ai/integrations/overview.md)
- [docs.guild.ai/api-reference/mcp](https://docs.guild.ai/api-reference/mcp.md)
- [docs.guild.ai/platform/credentials](https://docs.guild.ai/platform/credentials.md)
- [docs.guild.ai/platform/credential-policies](https://docs.guild.ai/platform/credential-policies.md)
- [docs.guild.ai/platform/security-architecture](https://docs.guild.ai/platform/security-architecture.md)
- [docs.guild.ai/llms.txt](https://docs.guild.ai/llms.txt)
- [docs.guild.ai/platform/triggers](https://docs.guild.ai/platform/triggers.md)
- [guild.ai/product/ai-governance](https://www.guild.ai/product/ai-governance)
- [docs.guild.ai/platform/evaluations](https://docs.guild.ai/platform/evaluations.md)
- [docs.guild.ai/platform/publish-to-agent-hub](https://docs.guild.ai/platform/publish-to-agent-hub.md)
- [docs.guild.ai/cli/commands](https://docs.guild.ai/cli/commands.md)
- [docs.guild.ai/reference/pricing](https://docs.guild.ai/reference/pricing.md)

## Other comparisons

- [Guild vs. Claude Managed Agents](https://www.guild.ai/compare/claude-managed-agents)
- [Guild vs. AWS](https://www.guild.ai/compare/aws)
- [Guild vs. LangChain](https://www.guild.ai/compare/langchain)
- [Guild vs. Runlayer](https://www.guild.ai/compare/runlayer)
- [Guild vs. Arcade](https://www.guild.ai/compare/arcade)
- [Guild vs. Vercel eve](https://www.guild.ai/compare/vercel-eve)
- [Guild vs. CrewAI](https://www.guild.ai/compare/crewai)
- [Guild vs. OpenAI Agents API](https://www.guild.ai/compare/openai-agents-api)
- [All comparisons](https://www.guild.ai/compare)
- [Book a demo](https://www.guild.ai/talk-to-sales)