# Celesto AI

Give your agents a Real Computer

Source: https://wheredoirunthis.com/products/celesto · JSON: https://wheredoirunthis.com/api/products/celesto · Read on 2026-10-12

## At a glance

- Type: Sandbox
- Kind: Open core
- Isolation: Virtual machine
- Free tier: No
- Starting price: Not disclosed
- GPU: No
- Regions: Not disclosed
- Coverage: 25 of 61 assessed capabilities supported
- Agent readiness: read 11 of 11 signals, drive 10 of 11 signals

## Links

- Homepage: https://celesto.ai/
- Docs: https://celesto.ai/docs
- Pricing: https://celesto.ai/pricing
- Repository: https://github.com/CelestoAI/celesto

## Pricing

- Model: Prepaid usage-metered running time, no monthly platform fee; top-ups above $2 (at least $2.01)
- Unit: Sandbox running hour at selected fixed compute shape
- Free tier: No

| Tier | Price | Includes | Notes |
| --- | --- | --- | --- |
| Trial | One-time $5 runtime credit for 30 days |  | No credit card required. Not a recurring free tier; top-up unlocks other sizes without ending original trial. [source](https://celesto.ai/pricing) |
| Micro | $0.015/running hour | 1 shared vCPU; 2 GB RAM; 20 GB root disk; Fixed root disk included with this running-time shape; not a separately provisioned GB-month credit. |  [source](https://celesto.ai/pricing) |
| Small | $0.020/running hour | 1 shared vCPU; 4 GB RAM; 25 GB root disk; Fixed root disk included with this running-time shape; not a separately provisioned GB-month credit. |  [source](https://celesto.ai/pricing) |
| Standard | $0.030/running hour | 2 shared vCPU; 4 GB RAM; 30 GB root disk; Fixed root disk included with this running-time shape; not a separately provisioned GB-month credit. |  [source](https://celesto.ai/pricing) |
| Memory | $0.040/running hour | 2 shared vCPU; 8 GB RAM; 40 GB root disk; Fixed root disk included with this running-time shape; not a separately provisioned GB-month credit. |  [source](https://celesto.ai/pricing) |
| Large | $0.080/running hour | 4 shared vCPU; 16 GB RAM; 60 GB root disk; Fixed root disk included with this running-time shape; not a separately provisioned GB-month credit. |  [source](https://celesto.ai/pricing) |
| XL | $0.160/running hour | 8 shared vCPU; 32 GB RAM; 100 GB root disk; Fixed root disk included with this running-time shape; not a separately provisioned GB-month credit. |  [source](https://celesto.ai/pricing) |

- Overages: No platform minimum; running time consumes prepaid credit, low balance stops sandboxes. Paid up to 10 running; stopped root disks retained free 30 days. No separate published workspace-storage or egress tariff found.
- Card required: No
- Source: https://celesto.ai/pricing

## Runtime

- **GPU & accelerators**: No — "Current machine sizes disclose shared vCPU/RAM/root disk only; no GPU workload offer found." [source](https://celesto.ai/pricing) (confidence: medium)
- **BYOC (Bring your own cloud)**: Not documented — "No current managed BYOC deployment flow found; local OSS self-hosting is separate." (source: Public website and documentation) (confidence: low)
- **Secrets management**: Yes — "Environment injection at boot/runtime; managed agent/provider credential connectivity documented. Secrets in guest env remain readable to workload code." [source](https://celesto.ai/docs/smolvm/guides/environment-variables) (confidence: high)
- **Storage options**: Included fixed root disks 20/25/30/40/60/100 GB by current size plus separately documented durable CelestoFS workspace filesystem — "CelestoFS petabyte-scale description is not an independently measured allocation or published object-storage bucket API." [source](https://celesto.ai/pricing) (confidence: high)
- **Supported workloads**: code-execution, computer-use-agents, browser-use-agents, coding-agents, long-running-agents, untrusted-code, preview-apps, github-actions-ci [source](https://celesto.ai/)
- **Persistence**: Durable root disk and CelestoFS workspace across stop/start; delete removes state; stopped root disks retained free for 30 days — "Pricing retention plus cloud persistence/CelestoFS docs." [source](https://celesto.ai/pricing) (confidence: high)
- **Data jurisdiction**: Not documented — "Corporate Delaware jurisdiction is known, hosted runtime region/residency is not disclosed." (source: Public website and documentation) (confidence: low)
- **Resource limits & quotas**: Current pricing: Micro 1 shared vCPU/2 GB RAM/20 GB root → XL 8 shared vCPU/32 GB RAM/100 GB root; paid up to 10 running sandboxes subject to host capacity; trial one retained Micro; publish at most 4 ports (1024–65535) — "Cloud resource docs disagree on shape names, RAM and max root disk; current pricing is the commercial source, effective provisioning not exercised." [source](https://celesto.ai/pricing) (confidence: medium)
- **SDK languages**: python, javascript, typescript [source](https://celesto.ai/docs/cloud/overview)
- **Network & egress controls**: Hosted open/off policy fixed at creation; no hosted domain/IP allowlists. Local core separately offers IPv4/CIDR restrictions with backend constraints. — ""Network Control currently supports these two settings. Domain and IP allowlists are not available." Offline hosted policy cannot combine with persistent_home=True." [source](https://celesto.ai/docs/cloud/features/network-control) (confidence: high)
- **Snapshots & forks**: Yes — "Homepage advertises forking, pause/resume and snapshots. Local core documents full/diff/disk checkpoints; hosted SDK docs primarily specify stop/start persistence, so exact cloud live-memory/fork API is not established." [source](https://celesto.ai/) (confidence: medium)
- **Deployment methods**: cli, sdk, api [source](https://celesto.ai/docs/cloud/overview)
- **Isolation**: Virtual machine — "Hosted offering discloses hardware-isolated microVMs but does not unambiguously identify its current hosted hypervisor. Record vm conservatively. The self-host core defaults to Firecracker on Linux, QEMU on macOS; libkrun is experimental. Do not import local-backend selection into a hosted guarantee." [source](https://celesto.ai/) (confidence: medium)

## Capabilities

### AI & agents

| Capability | Status | Evidence |
| --- | --- | --- |
| Agent skills | Partially supported | "Sibling Agentor has declarable tools/skills and dynamic loading; homepage advertises agent tools, but the hosted computer runtime itself is not a verified skills registry." [source](https://celesto.ai/docs/agentor/concepts/skills) |
| AI app builder | Not supported | "No prompt-to-app generation builder established. Running a third-party coding agent and exposing its result is not a first-party app-builder surface." [source](https://celesto.ai/) |
| Autonomous QA agent | Not supported | "Browser automation/test-capable sandboxes and OpenMuse coworker do not establish a first-party autonomous QA-testing product." [source](https://github.com/CelestoAI/celesto) |
| Built-in coding agent | Partially supported | "OpenMuse is a first-party open-source computer coworker preview; managed agents run user-defined agents. Neither is established as a shipped first-party code-authoring agent. Claude/Codex/etc. are third-party presets under preinstalled-coding-agent." [source](https://github.com/CelestoAI/celesto#built-with-celesto-openmuse) |
| BYOK for LLMs | Supported | ""ModelRequiresOwnKeyError" — this model can only run on your own provider key; provider credential connectivity is documented. Pi cloud integration explicitly keeps model credentials local." [source](https://celesto.ai/docs/cloud/managed-agents/reference) |
| Credential-injection proxy | Not supported | "Gatekeeper manages delegated end-user provider OAuth, not a documented on-wire phantom-token swap that prevents sandbox code seeing secrets. Pi remote tools keep keys local, which is also distinct from an injection proxy." [source](https://celesto.ai/docs/cloud/gatekeeper) |
| LLM gateway | Partially supported | "Managed agents select provider/model and report ProviderNotConnectedError/ModelRequiresOwnKeyError; Agentor supports provider routing. No standalone general-purpose multi-provider gateway API established for sandbox workloads." [source](https://celesto.ai/docs/cloud/managed-agents/reference) |
| MCP client | Partially supported | ""Define an agent, connect tools and MCP servers, and deploy." Detailed consuming-client implementation is Agentor CelestoMCPHub, a separate framework; hosted computer control is the MCP server side." [source](https://celesto.ai/) |
| MCP-server hosting | Partially supported | ""Ship agents ... through ... MCP servers" is advertised; a computer can run a Python web server and publish a port. Deployment API guide is being updated, and Agentor deployment docs now describe self-host ASGI; dedicated managed MCP-server deployment not established." [source](https://celesto.ai/) |
| MCP server | Supported | ""Hosted MCP" remote streamable HTTP URL https://mcp.celesto.ai/mcp; OAuth 2.1 DCR/PKCE. Unauthenticated GET returns 401 and resource_metadata header; protected resource and issuer metadata fetch successfully." [source](https://celesto.ai/agents.md) |
| Model inference API (first-party) | Not supported | "Managed agents select external provider models/keys; no first-party served model catalogue or standalone inference API established." [source](https://celesto.ai/docs/cloud/managed-agents/reference) |
| Preinstalled coding agent | Supported | "celesto codex/claude/pi/hermes/opencode start presets install/launch third-party agents in the local core. Hosted Pi extension is documented separately; not every local preset is asserted cloud-native." [source](https://github.com/CelestoAI/celesto#start-a-code-agent) |

### Networking & delivery

| Capability | Status | Evidence |
| --- | --- | --- |
| Automatic HTTPS | Supported | ""Port publishing creates a public HTTPS URL that forwards traffic to a port inside the computer"; preview/share app or API. Up to four ports; published URLs are public, add application auth for sensitive actions." [source](https://celesto.ai/docs/cloud/features/publish-ports) |
| Custom domains | Not supported | "Documented platform HTTPS preview URLs, not a bring-your-own-domain/DNS feature." [source](https://celesto.ai/docs/cloud/features/publish-ports) |
| Direct public IPv6 | Not supported | "No stable per-instance public IPv6/raw TCP/UDP endpoint documented; public access is a managed HTTPS port proxy." [source](https://celesto.ai/docs/cloud/features/publish-ports) |
| Edge KV store | Not applicable | "Sandbox compute/runtime, not a managed database or edge KV product; this data-layer primitive is not part of the exposed product." [source](https://celesto.ai/docs/cloud/overview) |
| Egress controls | Supported | ""Network Control currently supports these two settings. Domain and IP allowlists are not available." Hosted open/off policy at creation; offline policy excludes external persistent_home." [source](https://celesto.ai/docs/cloud/features/network-control) |
| Email send / receive | Not supported | "No platform-provided inbound/outbound workload mail service documented; Gmail/Calendar connector integrations are not a hosted mail primitive." [source](https://celesto.ai/docs/cloud/overview) |
| Public preview URLs | Supported | ""Port publishing creates a public HTTPS URL that forwards traffic to a port inside the computer"; preview/share app or API. Up to four ports; published URLs are public, add application auth for sensitive actions." [source](https://celesto.ai/docs/cloud/features/publish-ports) |
| Private networking | Not supported | "Sandbox isolation and outbound open/off policy do not establish a customer-private service mesh/network between that customer’s own hosted computers." [source](https://celesto.ai/docs/cloud/features/network-control) |
| Built-in reverse proxy | Supported | ""Port publishing creates a public HTTPS URL that forwards traffic to a port inside the computer"; preview/share app or API. Up to four ports; published URLs are public, add application auth for sensitive actions." [source](https://celesto.ai/docs/cloud/features/publish-ports) |

### Organisation & security

| Capability | Status | Evidence |
| --- | --- | --- |
| Built-in auth | Partially supported | "Gatekeeper delegates end-user OAuth access to third-party providers. Published sandbox app URLs are public and require user-supplied application authentication; arbitrary managed-agent end_user_id does not authenticate that user." [source](https://celesto.ai/docs/cloud/gatekeeper) |
| BYOC | Not supported | "Self-hostable Apache-2.0 runtime is available; no managed control-plane deployment into your cloud account established." [source](https://celesto.ai/) |
| Deploy into customers' VPCs | Not supported | "No SaaS-vendor deployment/management inside its customers’ VPCs documented; local runtime self-hosting is not a nested BYOC control plane." [source](https://celesto.ai/) |
| Secrets management | Supported | "Inject API keys, secrets and config at boot/runtime through the local Python API. Cloud managed-agent provider connections are separately documented. Guest environment variables are not on-wire secret redaction." [source](https://celesto.ai/docs/smolvm/guides/environment-variables) |
| Self-host / OSS core | Supported | ""Open-source core. Self-host or use our cloud." Apache-2.0 Celesto: "The microVM that boots every Celesto sandbox. Run it on your own infrastructure". Canonical repo CelestoAI/celesto, former SmolVM alias verified via GitHub API." [source](https://celesto.ai/) |
| Sharing | Supported | ""Port publishing creates a public HTTPS URL that forwards traffic to a port inside the computer"; preview/share app or API. Up to four ports; published URLs are public, add application auth for sensitive actions." [source](https://celesto.ai/docs/cloud/features/publish-ports) |
| SSO / SAML | Not supported | "No current Enterprise SSO/SAML offering established on pricing, docs or legal surface; the old June Enterprise claim was not carried forward." [source](https://celesto.ai/pricing) |
| Teams | Partially supported | "Cloud keys are organization-bound and full snapshot includes organization member APIs; source says internal endpoints are included. Current multi-user team plan/seats/invite UX was not established, so old Team-tier pass is not repeated." [source](https://github.com/CelestoAI/celesto/blob/main/openapi/cloud.json) |
| Workload identity | Not supported | "Organization API keys and human OAuth MCP authorization are not per-workload OIDC/SPIFFE identity federation; no such workload credential primitive documented." [source](https://celesto.ai/docs/cloud/authentication) |

### Developer access

| Capability | Status | Evidence |
| --- | --- | --- |
| CLI | Supported | "Official celesto CLI manages authentication, computers, commands, SSH and published ports." [source](https://celesto.ai/docs/cloud/cli) |
| E2B-compatible API | Not supported | "Own Computer/SDK API; no drop-in E2B endpoint/SDK compatibility claim established." [source](https://celesto.ai/docs/cloud/overview) |
| Filesystem API | Supported | "Local SDK bridge provides computer/sandbox/browser-session files endpoints and README computer.files.write; non-SSH local API path. Hosted Computers docs primarily expose command execution; do not presume a complete remote hosted file API." [source](https://github.com/CelestoAI/celesto/blob/main/ts/celesto-local-server.openapi.json) |
| Git deploy | Not supported | "Current deployment API is being updated; clone/run/publish inside a computer is not automatic git-push deployment. Managed CI runners are separate CI compute, not application git deploy." [source](https://celesto.ai/docs/cloud/deployments) |
| HTTP data API | Not applicable | "Sandbox compute/runtime, not a managed database or edge KV product; this data-layer primitive is not part of the exposed product." [source](https://celesto.ai/docs/cloud/overview) |
| OpenAPI spec | Supported | "Fetchable valid OpenAPI 3.1.0 cloud snapshot (186 paths); separate local SDK bridge spec 3.1.0 (19 paths). Cloud source note identifies backend revision and warns internal endpoints are included; only handwritten public Computer wrapper supported." [source](https://github.com/CelestoAI/celesto/blob/main/openapi/cloud.json) |
| Private container registry | Not supported | "Image builders/rootfs images exist, not a platform-hosted private container registry." [source](https://celesto.ai/docs/smolvm/api/imagebuilder) |
| Public API | Supported | "Documented Computers API create/get/list/commands/ports/terminal/lifecycle plus managed-agent namespaces." [source](https://celesto.ai/docs/cloud/computers) |
| SDK | Supported | "Official Python celesto and TypeScript/JavaScript @celestoai/sdk; shared Computer resource API." [source](https://celesto.ai/docs/cloud/overview) |
| SSH access | Supported | "celesto computer ssh COMPUTER_ID --cloud; local core has computer ssh and faster shell/vsock paths." [source](https://celesto.ai/) |

### Compute & storage

| Capability | Status | Evidence |
| --- | --- | --- |
| Cloud-init / ignition config | Not supported | "Custom VM images and environment injection exist; no supported cloud-init/Ignition user-data boot interface established." [source](https://celesto.ai/docs/smolvm/installation) |
| Desktop / computer-use sandbox | Supported | ""A real computer for every agent" with full browser; Browser Use headless or noVNC streamed live. Local Celesto.browser()/computer() and desktop APIs are documented; macOS/Windows guest previews remain local-core scoped." [source](https://celesto.ai/) |
| GPU compute | Not supported | "Current six sizes disclose shared CPU/RAM/root disk only; no GPU workload offering found." [source](https://celesto.ai/pricing) |
| Live-state (memory) snapshots | Supported | "Local full snapshots capture CPU state, memory and disk, restoring running process state; disk-only snapshots reboot. Firecracker/QEMU Linux isolated-disk scope; Windows/workspace mounts/extra drives excluded. Exact hosted cloud memory API not independently established." [source](https://celesto.ai/docs/smolvm/features/snapshots) |
| Multi-region | Not supported | "No public selectable hosted region list or placement parameter established; company HQ and subprocessor geographies are not workload regions." [source](https://celesto.ai/docs/cloud/overview) |
| Object storage | Not supported | "CelestoFS is a durable mounted workspace filesystem over object storage; no user bucket/S3/object CRUD/presigned object-storage offering documented." [source](https://celesto.ai/docs/cloud/features/petabyte-storage) |
| Persistent disk | Supported | "Durable root disk and workspace state across stop/start; deleting removes files. Current pricing retains stopped root disks free for 30 days." [source](https://celesto.ai/docs/cloud/features/persistence) |
| Sandbox templates / custom images | Supported | "Hosted scratch/coding-agent templates; local ImageBuilder/DockerRootfsBuilder builds custom VM rootfs images. Custom customer image build scope is clearly documented for OSS, not presumed cloud registry support." [source](https://celesto.ai/docs/cloud/features/resources) |
| Scale to zero | Supported | ""Sandboxes stop when it is too low" and "Stopped root disks are retained free for 30 days"; charge is running time, no monthly platform fee. Explicit stop/start rather than guaranteed automatic idle timeout." [source](https://celesto.ai/pricing) |
| Selectable isolation runtime | Supported | "Local backend option accepts auto, firecracker, qemu and experimental libkrun; Firecracker/QEMU are stable. Hosted per-computer hypervisor selection is not established." [source](https://celesto.ai/docs/smolvm/concepts/backends) |
| Snapshots / forking | Supported | ""Branch a running sandbox into parallel copies" and pause/resume/snapshots advertised. Detailed local full/diff/disk snapshot docs; hosted SDK fork API not independently established." [source](https://celesto.ai/) |

### Data

| Capability | Status | Evidence |
| --- | --- | --- |
| Database branching | Not applicable | "Sandbox compute/runtime, not a managed database or edge KV product; this data-layer primitive is not part of the exposed product." [source](https://celesto.ai/docs/cloud/overview) |
| Ephemeral instant DB (no-signup) | Not applicable | "Sandbox compute/runtime, not a managed database or edge KV product; this data-layer primitive is not part of the exposed product." [source](https://celesto.ai/docs/cloud/overview) |
| First-party database | Not supported | "No user-provisioned managed database offering documented; internal SQLite state or stored agent transcripts are not a first-party user database." [source](https://celesto.ai/docs/cloud/overview) |
| Vector search | Not applicable | "Sandbox compute/runtime, not a managed database or edge KV product; this data-layer primitive is not part of the exposed product." [source](https://celesto.ai/docs/cloud/overview) |

### Platform & operations

| Capability | Status | Evidence |
| --- | --- | --- |
| Managed CI/CD runners | Supported | ""GitHub workflows. Celesto compute." Install Celesto CI and choose Micro through XL; fresh Ubuntu Linux x64 per runner. No Windows/macOS, Docker actions, job containers or service containers." [source](https://celesto.ai/ci) |
| Built-in observability | Supported | "Hosted dashboard receives opt-in Agentor traces of prompts/tool calls/timings/tokens. Current Agentor docs say tracing defaults off, despite homepage automatic-tracing wording; managed agents retain run events/audit history." [source](https://celesto.ai/docs/agentor/tracing) |
| Preview environments | Not supported | "Shareable app preview URL exists, not a complete per-PR compute/addon environment with automatic teardown." [source](https://celesto.ai/docs/cloud/features/publish-ports) |
| Scheduled tasks | Not supported | "No documented platform cron/scheduled sandbox task primitive found; installing cron in the VM would not be first-party scheduling." [source](https://celesto.ai/docs/cloud/overview) |
| Project template marketplace | Not supported | "Compute scratch/coding-agent images are sandbox templates, not a project-starter app marketplace." [source](https://celesto.ai/docs/cloud/features/resources) |
| Durable workflows | Partially supported | "Sibling Agentor saves steps to disk to resume/fork an agent run. Cloud computer stop/start persistence and managed-agent stored transcripts are not a general hosted durable-workflow engine." [source](https://celesto.ai/docs/agentor/durable-runs) |

## Agent readiness

### Read: 11 of 11 signals present

11/11 signals (7 full, 4 partial)

- Llms txt: pass — "/llms.txt 200 substantive site index and /docs/llms.txt 200 rich page index (33811 bytes)."
- Llms full txt: pass — "/docs/llms-full.txt 200 comprehensive 1004137-byte corpus; date-prefixed POINTER retained. Apex /llms-full.txt 404 does not negate docs copy."
- Robots AI policy: partial — "robots.txt names search/social crawlers plus User-agent:* Allow:/; silent about named AI UAs, default-allow."
- JSON LD: partial — "Homepage has Organization/WebSite/SoftwareApplication JSON-LD graph with name/description, but no offers in the SoftwareApplication record."
- Sitemap: pass — "Root sitemap 12 URLs; docs XML sitemap 19417 bytes; robots also points to blog sitemap index, successfully rendered after direct 403."
- Meta quality: pass — "Descriptive title, meta description, OG title/description/url/site/image and Twitter metadata in raw server HTML."
- Markdown negotiation: pass — "Accept:text/markdown on /docs/cloud/overview returns 200 text/markdown (6156 bytes); homepage remains HTML. Per-page docs .md variants also work."
- Content efficiency: partial — "HTMLParser excludes script/style: 7928 readable-text UTF-8 bytes / 188029 HTML bytes = 4.216%, rubric 1–5% half credit. Homepage is no longer a tiny empty JS shell."
- Docs discoverable: pass — "Header Docs and hero View docs link to https://celesto.ai/docs; old docs.celesto.ai redirects to the current docs path."
- Pricing machine readable: pass — "Dedicated /pricing raw HTML contains parseable six-row CPU/RAM/root-disk/hourly-price table, prepaid/trial/retention copy."
- Anti bot: partial — "Main pages/docs/specs directly fetch; blog and blog sitemap direct requests return Cloudflare error 1010/403, both rescued by sequential foxcrawl 200 (blocked:false). Intermittent/some-path public challenge, not persistent missing content."

### Drive: 10 of 11 signals present

10/11 signals (7 full, 3 partial)

- Public API: pass — "Documented Computers API and managed agents namespaces; cloud SDK/API surface public."
- Openapi spec: pass — "Public GitHub cloud OpenAPI 3.1.0 snapshot with 186 paths and local bridge 19-path OpenAPI spec. API host /openapi.json 404; source repo published specs credited, not a false absent."
- MCP server: pass — "agents.md documents https://mcp.celesto.ai/mcp with OAuth DCR/PKCE. 401 WWW-Authenticate resource_metadata resolves to public protected-resource metadata and API-host OAuth issuer discovery 200."
- SDK packages: pass — "Python celesto and TypeScript/JavaScript @celestoai/sdk documented official clients."
- Auth docs: pass — "Credential lookup order, organization scoping, API-key login/store/logout and OAuth MCP issuer/auth/token/DCR/scope discovery documented."
- Agent self serve creds: partial — "Settings > Security key creation documented; MCP OAuth still requires account browser login. DCR metadata is not proof an autonomous agent can register without a human."
- Webhooks: fail — "No documented outbound platform webhook subscription with signing. Exposing webhook receivers and internal Stripe/email/Telegram webhook endpoints do not qualify."
- CLI tool: pass — "Official celesto CLI local/cloud creates, executes, SSHs and manages computers/ports."
- Error model: pass — "Documented SDK HTTP error classes/status/data/requestId and managed-agent machine-readable refusal codes plus OpenAPI JSON schemas. Real unauthenticated API/MCP response includes detail/status_code."
- Rate limit docs: partial — "CelestoRateLimitError described for too many requests; managed-session conflict documents Retry-After hint. No numeric general API quota plus rate-limit response headers established."
- Agent config: partial — "Public /agents.md contains MCP tools/client configuration. Repository AGENTS.md is development guidance, not an independent hosted integration discovery surface; /.well-known/agent returns 404."

## Coding harnesses

- Claude Code (Anthropic) — "celesto claude start" [source](https://github.com/CelestoAI/celesto#start-a-code-agent)
- Codex CLI (OpenAI) — "celesto codex start" [source](https://github.com/CelestoAI/celesto#start-a-code-agent)
- OpenCode (opencode) — "celesto opencode start" [source](https://github.com/CelestoAI/celesto#start-a-code-agent)
- Pi (earendil-works) — "celesto pi start" [source](https://github.com/CelestoAI/celesto#start-a-code-agent)
- Hermes Agent (Nous Research) — "celesto hermes start" [source](https://github.com/CelestoAI/celesto#start-a-code-agent)

Plausible compatibility (derived, not stored):

- Aider
- Amp
- Antigravity
- Devin
- Droid
- Gemini CLI
- Goose
- Kilo Code
- Kiro CLI
- Qwen Code

## Compliance & regions

- Regions note: No user-selectable hosted region list or runtime data-residency commitment found. Delaware corporate registration and London headquarters are company facts, not cloud regions. Subprocessor locations are not user workload placement. Self-host placement is operator-controlled.

## Open source

- Repository: https://github.com/CelestoAI/celesto
- License: Apache-2.0
- Stars: 1028
- Forks: 84
- Self-host methods: package-manager, source-build

## Company & presence

- Headquarters: London, United Kingdom
- Privacy policy: https://celesto.ai/legal/privacy-policy
- Terms of service: https://celesto.ai/legal/terms
- GitHub (core): https://github.com/CelestoAI/celesto
- GitHub (org): https://github.com/CelestoAI
- Docs: https://celesto.ai/docs
- X/Twitter: https://x.com/CelestoAI
- Discord: https://discord.gg/KNb5UkrAmm
- LinkedIn: https://www.linkedin.com/company/celestoai
- Blog: https://celesto.ai/blog/
- Managed GitHub Actions CI: https://celesto.ai/ci

## Nearby in the atlas

- [Agent Safehouse](/products/agent-safehouse.md)
- [Agent Sandbox](/products/agent-sandbox.md)
- [Agent Substrate](/products/agent-substrate.md)
- [Agent37](/products/agent37.md)
- [AgentNest](/products/agentnest.md)
- [Anthropic Sandbox Runtime (srt)](/products/sandbox-runtime.md)

## Method

Facts come from the product's own public pages. Each claim carries its evidence: a quote and the page it came from. We count capabilities rather than grade products, and "Not assessed" means we have not checked, never that the product lacks it.