> ## Documentation Index > Fetch the complete documentation index at: https://docs.orgo.ai/llms.txt > Use this file to discover all available pages before exploring further. # Apply secrets to many computers > Write vault secrets onto every computer in a workspace, or a list of computers, in one call. Take them off, or preview what either would change. Applies vault secrets to many running Linux computers in one call. With `remove: true` it takes them off instead, and with `dry_run: true` it changes nothing and says what each computer would get. Each computer gets what [Apply secrets to a computer](https://docs.orgo.ai/api-reference/computers/sync-secrets.md) gives it: the same files, the same agent env files, and the same `report`. Values are never returned. > [!NOTE] > > This call never wakes a computer. A computer that is asleep, in storage or on its host, is listed as `not_changed`. Start or wake it, then send its id again. ## Body parameters **names** `string[]` *required* 1 to 50 environment variable names from your vault: letters, digits, and underscores, not starting with a digit, up to 64 characters each. Duplicates are ignored. `ORGO_API_KEY` returns `422`. **computer\_ids** `string[] | null` 1 to 100 computers, each a computer UUID or `instance_id`. Orgo sends to exactly these computers. Duplicates are ignored. `null` is the same as leaving it out. **workspace\_id** `string | null` A workspace UUID, in lowercase as Orgo returns it. Without `computer_ids`, every computer in this workspace is in scope. With `computer_ids`, it does not choose computers, but it must still be a workspace you can change. Any other string, an id in capitals included, returns `403`. `null` is the same as leaving it out. **remove** `boolean` (default: false) `true` takes the names off the computers instead of writing them, as described in [Remove secrets from a computer](https://docs.orgo.ai/api-reference/computers/sync-secrets.md#remove-secrets-from-a-computer). **dry\_run** `boolean` (default: false) `true` changes nothing. Each computer answers what an apply, or with `remove: true` a removal, would change there, as described in [Preview what would change](https://docs.orgo.ai/api-reference/computers/sync-secrets.md#preview-what-would-change). `remove` and `dry_run` are each `true`, `false`, or left out. Any other value returns `400`. Fields this endpoint does not take, such as `wake`, are ignored. ## Which computers | You send | Computers in scope | | - | - | | `computer_ids` | Exactly those. A computer you can't reach or can't change is a `failed` or `not_changed` result, not a refusal of the whole call. | | `workspace_id` | Every computer in that workspace. Running Linux computers get every name. The rest are `not_changed`, with why. | | Neither | The computers where each name is saved, described below. | With neither, each name goes where it is saved: * An account secret goes to every workspace you own. * A secret saved only for a workspace goes to that workspace, whoever owns it. * A workspace someone else owns does not get account secrets, because its other members, and the agents they talk to, could read what you send. Name it in `workspace_id`, or its computers in `computer_ids`, to include it. Unless every name you send that has an account value also has its own value saved for that workspace, none of those names go there, not even the workspace's own values. Names saved only for that workspace still go there. When nothing goes there, its running computers are listed as `not_changed`, with how to include them. * With `remove: true`, an account secret is taken off every workspace you own, except where a workspace has its own secret with that name. Its computers got the workspace's value, so the name stays there. When that leaves nothing to remove in a workspace, its computers are listed as `not_changed`. * A removal leaves out a workspace someone else owns in the same way, because it takes the name off whatever its value, so it would also take another member's value of that name. Name the workspace or its computers to include it. A secret saved only for that workspace is still taken off there. * A workspace you can only view is never changed. When a secret saved only for that workspace names it, its computers are listed as `not_changed`, saying you can only view it. Otherwise it is not listed. * Client workspaces you created under Invoicing are left out. Name them in `workspace_id` or `computer_ids`. * A name that is in none of your vaults returns `404` with `code: "secret_not_found"`. Where a workspace has its own value for a name and Orgo sends that name there, its computers get that value rather than the account's. ## How the call runs Orgo works on four computers at a time, and never wakes one. A computer whose secrets are too large to send in one call is sent them in halves. No computer starts after 45 seconds, and a computer sent its secrets in halves starts no next half after that. Computers already under way finish, so the answer can take longer than 45 seconds. Orgo starts no step on a computer that could still be running 90 seconds into the call. A removal leaves out its second step on `/root/.env` when time is short, and that computer's `report` says the line stays. The computers that did not start are in `remaining`, and `next` holds the calls that finish the job. ## Response **dry\_run** `boolean` Whether this was a preview. **names** `string[]` The names sent, without duplicates. **results** `object[]` One entry per computer, in the order of the computers in scope, followed by the computers found `not_changed` while choosing them. **properties** **computer\_id** `string` The computer's UUID when Orgo could read the computer, else the id as you sent it. **name** `string | null` The computer's name. `null` when it is not known, such as for a computer that was not found. **workspace\_id** `string | null` The computer's workspace. `null` when it is not known. **outcome** `string` `applied`, `removed`, `previewed`, `not_changed`, or `failed`. **reason** `string` On `not_changed` and `failed`: why, as a sentence you can show people. **report** `object | null` On `applied`, `removed`, and `previewed`: what the computer reported, described in [The report](https://docs.orgo.ai/api-reference/computers/sync-secrets.md#the-report). `null` when no report came back. **remaining** `string[]` The computers that did not start before the time ran out: the ids as you sent them in `computer_ids`, or computer UUIDs otherwise. Empty when every computer ran. **next** `object[]` The calls that finish the job: the computers in `remaining`, each call with the names those computers were due. Send each one to this endpoint as it is. Empty when `remaining` is. **properties** **names** `string[]` The names these computers were due. With neither `computer_ids` nor `workspace_id`, a workspace can be due fewer names than you sent. **computer\_ids** `string[]` Up to 100 of the computers in `remaining`. **remove** `boolean` `true` when this call was a removal. Left out otherwise. **dry\_run** `boolean` `true` when this call was a preview. Left out otherwise. The answer carries `Cache-Control: no-store`. ### Outcomes | `outcome` | Meaning | | - | - | | `applied` | The names were written. `report` says what each agent got. | | `removed` | The removal ran. `report.names` says, for each name, whether a copy was taken out or stays. | | `previewed` | `dry_run: true`. Nothing changed, and `report` says what would. | | `not_changed` | The computer was not sent to, or answered that it can't be changed now. `reason` says why. | | `failed` | The computer could not be changed. `reason` says why. | Common reasons: | `outcome` | `reason` | When | | - | - | - | | `failed` | `Computer not found.` | No such computer, or one you are not a member of. The two read the same, so the answer does not tell which ids exist. | | `failed` | `You can view this workspace but not change its computers.` | You can only view the computer's workspace, and you named it in `computer_ids`. | | `failed` | `A requested secret is not available in your vault for this workspace.` | A name is not in your vault for the computer's workspace. Apply and its preview only. | | `failed` | `Could not apply secrets. Check the computer connection and try again.` | The computer did not answer. It says `remove` for a removal and `check` for a preview. | | `failed` | `This computer’s secrets file could not be read, so nothing was changed.` | `/root/.orgo-secrets.json` on the computer does not parse. | | `failed` | `These secrets are too large to apply in one request. Apply fewer at a time.` | One name's value alone is too large to send to the computer. | | `failed` | `Couldn’t check this computer. Try again.` | Orgo could not look the computer up. | | `failed` | Why Orgo refused the computer, such as `Choose a plan to continue.` | The computer's trial or payment has lapsed, or Orgo locked it. | | `not_changed` | `Asleep. Not changed.` | The computer is suspended, in storage or on its host, or went to sleep while Orgo reached it. Start or wake it, then send its id again. | | `not_changed` | `Stopped. Not changed.`, `Not reachable. Not changed.`, or `Not running. Not changed.` | The computer was not running when Orgo chose the computers. | | `not_changed` | `Start the computer before applying secrets.` | A computer you named in `computer_ids` is not running, or is suspended on its host. | | `not_changed` | `Windows computers can’t get secrets yet.` (or Mac, Android, iPhone, or `This kind of computers can’t get secrets yet.` for any other system) | Only Linux computers get secrets. | | `not_changed` | `Applying secrets currently requires a Linux computer.` | A computer you named in `computer_ids` is not a Linux computer. | | `not_changed` | `Someone else owns this workspace. Name it in workspace_id or computer_ids to include it.` | The default scope left out a workspace someone else owns. | | `not_changed` | `This workspace has its own value. Not changed.` | A removal with the default scope left out a workspace with its own secret of that name. | | `not_changed` | `You can view this workspace but not change its computers.` | The default scope found a secret saved for a workspace you can only view. | ## Continue with remaining When `remaining` is not empty, send each body in `next` to this endpoint as it is, and do the same with each answer until `next` is empty. Running the same apply on a computer twice is safe: the values are merged, not added twice. Each body in `next` names the computers in `computer_ids`, up to 100 a call, which is the most `computer_ids` takes. Its `names` are what those computers were due. With the default scope that can be fewer than you sent: a workspace with its own value for a name keeps it on a removal, and a secret saved for one workspace goes to that workspace only. Sending every name to those computers instead fails where a name is not in your vault for the computer's workspace, and a removal would take off a name the workspace keeps its own value for. ## Errors | Status | `code` | When | | - | - | - | | `400` | none | The body is not valid: `names must contain 1 to 50 environment variable names (letters, numbers, underscores; no leading number). computer_ids holds 1 to 100 computer ids, workspace_id is a workspace id, and remove and dry_run are true or false.` A `workspace_id` that is not a string returns this too. | | `401` | none | No usable credential. | | `401`, `403` | `key_expired`, `key_permission`, `computer_key`, `computer_scope`, and others | The key itself was refused before the call ran. A restricted key needs Secrets write. A computer's own key and a key for one computer can't call this endpoint. Listed in [Codes any endpoint can return to a key](https://docs.orgo.ai/api-reference/api-keys/manage.md#codes-any-endpoint-can-return-to-a-key). | | `403` | `workspace_scope_mismatch` | A key scoped to one workspace: `This endpoint requires an account-wide credential (workspace_scope_mismatch).` Use a key for your whole account. | | `403` | none | `workspace_id` is not the id of a workspace you are a member of, such as a string that is not a UUID or a UUID in capitals: `You do not have access to this workspace.` | | `403` | `workspace_read_only` | `workspace_id` names a workspace you can only view: `This workspace is view-only. Ask the owner for write access (workspace_read_only).` | | `404` | `secret_not_found` | With neither `computer_ids` nor `workspace_id`, a name is in none of your vaults: `NAME is not in your vault.` For a removal: `NAME is not in your vault, so Orgo can’t tell which computers have it. Name them in computer_ids, or a workspace in workspace_id.` | | `422` | `reserved_name` | A name is `ORGO_API_KEY`: `ORGO_API_KEY is each computer’s own key. Orgo sets it, so it can’t be applied or removed here.` | | `500` | none | An unexpected failure, such as a database error while Orgo chooses the computers: `Could not apply secrets. Try again.`, `Could not remove secrets. Try again.`, or for a preview `Could not check your computers. Try again.` | | `503` | none | Orgo could not verify the credential because of a server-side fault: `Service temporarily unavailable. …`, with `Retry-After: 5`. Retry after that delay; do not rotate the key. | A refusal for one computer never fails the call. It is that computer's `failed` or `not_changed` result. ## Example The cURL call previews what applying `ANTHROPIC_API_KEY` would change on every computer in a workspace. The Python one applies it, then sends each body in `next` until every computer has run. Set `$WORKSPACE_ID` to the workspace's id. ```bash cURL theme={null} curl -X POST https://www.orgo.ai/api/secrets/apply \ -H "Authorization: Bearer $ORGO_API_KEY" \ -H "Content-Type: application/json" \ -d '{"names": ["ANTHROPIC_API_KEY"], "workspace_id": "'"$WORKSPACE_ID"'", "dry_run": true}' ``` ```python Python theme={null} import os import requests url = "https://www.orgo.ai/api/secrets/apply" headers = {"Authorization": f"Bearer {os.environ['ORGO_API_KEY']}"} # Apply to every computer in the workspace, then to any the time ran out for. calls = [{"names": ["ANTHROPIC_API_KEY"], "workspace_id": os.environ["WORKSPACE_ID"]}] while calls: response = requests.post(url, headers=headers, json=calls.pop(0)) response.raise_for_status() answer = response.json() for result in answer["results"]: print(result["name"], result["outcome"], result.get("reason", "")) calls.extend(answer["next"]) ``` ### Response ```json theme={null} { "dry_run": true, "names": ["ANTHROPIC_API_KEY"], "results": [ { "computer_id": "a3bb189e-8bf9-3888-9912-ace4e6543002", "name": "agent-1", "workspace_id": "6f1c2e7a-4b9d-4e2f-9a51-0c3d8e7b2f10", "outcome": "previewed", "report": { "ok": true, "targets": [ { "id": "hermes", "status": "would_update", "names": ["ANTHROPIC_API_KEY"] }, { "id": "openclaw", "status": "would_update", "names": ["ANTHROPIC_API_KEY"], "restart": "openclaw-gateway" }, { "id": "claude-code", "status": "shell" } ] } }, { "computer_id": "c41f7a2e-1d3b-4c8e-b6a9-5e2d0f9c7b31", "name": "agent-2", "workspace_id": "6f1c2e7a-4b9d-4e2f-9a51-0c3d8e7b2f10", "outcome": "not_changed", "reason": "Asleep. Not changed." } ], "remaining": [], "next": [] } ``` ## OpenAPI ````yaml POST /secrets/apply openapi: 3.1.0 info: title: Orgo API description: >- Launch cloud computers that AI agents can control and interact with. Create workspaces, provision computers, and control them programmatically. version: 2.0.0 contact: name: Orgo Support email: contact@orgo.ai url: https://orgo.ai servers: - url: https://www.orgo.ai/api description: Production security: - bearerAuth: [] tags: - name: Support description: >- Orgo’s support agent, for programs: ask how to do something in Orgo and get an answer with full URLs and copyable commands. - name: Account description: >- Account capacity: how many computers an account may run, and adding or giving back more. - name: Agent accounts description: >- Accounts an AI agent makes for itself, and the claim link a person uses to take one over. - name: Clients description: >- Run Orgo for your clients from your own app: a workspace and scoped key each, billed to you or to them. - name: Workspaces description: Organize computers into named workspaces - name: Computers description: Provision and manage virtual computers - name: Computer Lifecycle description: Start, stop, and restart computers - name: Computer Actions description: Control mouse, keyboard, and execute commands - name: Screens description: >- More than one desktop on a single computer. Each screen is its own X server with its own cursor and window manager, so an agent working on one cannot disturb another. - name: Files description: Upload and download files - name: Templates description: Author, build, and launch reproducible computers from templates - name: Chat description: Run an agent on a computer through OpenAI-compatible chat completions. paths: /secrets/apply: post: tags: - Computers summary: Apply secrets to many computers description: >- Applies vault secrets to many running Linux computers in one call, takes them off with `remove: true`, or answers what either would change with `dry_run: true`. With `computer_ids`: exactly those computers. With `workspace_id`: every computer in that workspace. With neither: each name where it is saved, a workspace secret to its workspace and an account secret to every workspace you own. A workspace someone else owns gets account secrets, or loses them on a removal, only when named. Each computer gets what `POST /computers/{id}/secrets/sync` gives it. Four computers at a time. Never wakes a computer. No computer starts after 45 seconds, and no step on a computer starts that could still be running 90 seconds in. The computers that did not start are in `remaining`, and `next` holds the calls that finish the job, each ready to send as it is. A computer that can't be changed is a `failed` or `not_changed` result, never a refusal of the whole call. Values are never returned. operationId: applySecretsToComputers requestBody: required: true content: application/json: schema: type: object required: - names properties: names: type: array minItems: 1 maxItems: 50 items: type: string pattern: ^[A-Za-z_][A-Za-z0-9_]{0,63}$ description: >- An environment variable name: letters, digits, and underscores, not starting with a digit, up to 64 characters. description: >- The vault names to apply, remove, or preview. Duplicates are ignored. `ORGO_API_KEY` answers `422`. computer_ids: type: - array - 'null' minItems: 1 maxItems: 100 items: type: string pattern: ^[A-Za-z0-9][A-Za-z0-9._-]{0,127}$ description: A computer UUID or `instance_id`. description: >- Exactly these computers. Duplicates are ignored. `null` is the same as leaving it out. workspace_id: type: - string - 'null' format: uuid description: >- Without `computer_ids`: every computer in this workspace. With `computer_ids` it does not choose computers, but it must still be a workspace you can change. A string that is not the id of a workspace you are a member of, a UUID in capitals included, answers `403`. `null` is the same as leaving it out. remove: type: boolean default: false description: >- `true` takes the names off the computers instead of writing them. dry_run: type: boolean default: false description: >- `true` changes nothing. Each computer answers what an apply, or with `remove: true` a removal, would change there. examples: default: summary: Where Settings would apply value: names: - ANTHROPIC_API_KEY workspace: summary: Every computer in a workspace value: names: - ANTHROPIC_API_KEY workspace_id: 6f1c2e7a-4b9d-4e2f-9a51-0c3d8e7b2f10 computers-preview: summary: Preview on chosen computers value: names: - ANTHROPIC_API_KEY computer_ids: - a3bb189e-8bf9-3888-9912-ace4e6543002 - c41f7a2e-1d3b-4c8e-b6a9-5e2d0f9c7b31 dry_run: true remove: summary: Remove value: names: - ANTHROPIC_API_KEY remove: true responses: '200': description: >- Each computer's outcome. The answer carries `Cache-Control: no-store`. content: application/json: schema: type: object required: - dry_run - names - results - remaining - next properties: dry_run: type: boolean names: type: array items: type: string description: The names sent, without duplicates. results: type: array items: $ref: '#/components/schemas/SecretsApplyResult' description: >- One entry per computer, in the order of the computers in scope, followed by the computers found `not_changed` while choosing them. remaining: type: array items: type: string description: >- The computers that did not start before the time ran out: the ids as sent in `computer_ids`, or computer UUIDs otherwise. `next` holds the calls that finish them. next: type: array description: >- The calls that finish the job: the computers in `remaining`, grouped by the names each was due, up to 100 a call. Send each one to this endpoint as it is. With neither `computer_ids` nor `workspace_id`, a workspace can be due fewer names than were sent. Empty when `remaining` is. items: type: object required: - names - computer_ids properties: names: type: array items: type: string description: The names these computers were due. computer_ids: type: array minItems: 1 maxItems: 100 items: type: string description: Up to 100 of the computers in `remaining`. remove: type: boolean enum: - true description: Present, `true`, when the call was a removal. dry_run: type: boolean enum: - true description: Present, `true`, when the call was a preview. example: dry_run: false names: - ANTHROPIC_API_KEY results: - computer_id: a3bb189e-8bf9-3888-9912-ace4e6543002 name: agent-1 workspace_id: 6f1c2e7a-4b9d-4e2f-9a51-0c3d8e7b2f10 outcome: applied report: ok: true targets: - id: hermes status: updated names: - ANTHROPIC_API_KEY - id: claude-code status: shell - computer_id: c41f7a2e-1d3b-4c8e-b6a9-5e2d0f9c7b31 name: agent-2 workspace_id: 6f1c2e7a-4b9d-4e2f-9a51-0c3d8e7b2f10 outcome: not_changed reason: Asleep. Not changed. remaining: [] next: [] '400': description: >- `names` is not a list of 1 to 50 valid environment variable names, `computer_ids` is not a list of 1 to 100 computer ids, `workspace_id` is not a string, or is empty, or `remove` or `dry_run` is not a boolean. content: application/json: schema: $ref: '#/components/schemas/Error' example: error: >- names must contain 1 to 50 environment variable names (letters, numbers, underscores; no leading number). computer_ids holds 1 to 100 computer ids, workspace_id is a workspace id, and remove and dry_run are true or false. '401': $ref: '#/components/responses/Unauthorized' '403': description: >- The key may not call this endpoint: a restricted key without Secrets write (`key_permission`), a computer's own key (`computer_key`), a key for one computer (`computer_scope`), or a key scoped to one workspace (`workspace_scope_mismatch`). Or `workspace_id` is not the id of a workspace you are a member of (a string that is not a UUID, or a UUID in capitals, included), or names one you can only view (`workspace_read_only`). content: application/json: schema: $ref: '#/components/schemas/Error' examples: workspace-key: summary: The API key is workspace-scoped value: error: >- This endpoint requires an account-wide credential (workspace_scope_mismatch). code: workspace_scope_mismatch no-access: summary: Not a member of workspace_id value: error: You do not have access to this workspace. view-only: summary: workspace_id is view only value: error: >- This workspace is view-only. Ask the owner for write access (workspace_read_only). code: workspace_read_only '404': description: >- With neither `computer_ids` nor `workspace_id`, a name is in none of your vaults. content: application/json: schema: $ref: '#/components/schemas/Error' examples: apply: summary: Apply value: error: ANTHROPIC_API_KEY is not in your vault. code: secret_not_found remove: summary: Remove value: error: >- ANTHROPIC_API_KEY is not in your vault, so Orgo can’t tell which computers have it. Name them in computer_ids, or a workspace in workspace_id. code: secret_not_found '422': description: A name is `ORGO_API_KEY`, which holds each computer's own key. content: application/json: schema: $ref: '#/components/schemas/Error' example: error: >- ORGO_API_KEY is each computer’s own key. Orgo sets it, so it can’t be applied or removed here. code: reserved_name '500': description: >- An unexpected failure, such as a database error while Orgo chooses the computers. content: application/json: schema: $ref: '#/components/schemas/Error' examples: apply: summary: Apply value: error: Could not apply secrets. Try again. remove: summary: Remove value: error: Could not remove secrets. Try again. preview: summary: Preview value: error: Could not check your computers. Try again. '503': $ref: '#/components/responses/AuthUnavailable' components: schemas: SecretsApplyResult: type: object required: - computer_id - name - workspace_id - outcome properties: computer_id: type: string description: >- The computer's UUID when Orgo could read the computer, else the id as you sent it. name: type: - string - 'null' description: >- The computer's name. `null` when it is not known, such as for a computer that was not found. workspace_id: type: - string - 'null' description: The computer's workspace. `null` when it is not known. outcome: type: string enum: - applied - removed - previewed - not_changed - failed reason: type: string description: >- On `not_changed` and `failed`: why, as a sentence you can show people, such as `Asleep. Not changed.` or `Computer not found.` report: $ref: '#/components/schemas/SecretsReport' description: >- On `applied`, `removed`, and `previewed`: what the computer reported. `null` when no report came back. Error: type: object description: >- The base error body. Every failure carries `error`; individual endpoints add the fields named in the schemas below. required: - error properties: error: type: string description: Human-readable message. example: Access denied code: type: string description: >- Machine-readable reason. Present on the failures that define one, absent otherwise. SecretsReport: type: - object - 'null' description: >- What happened, or would happen, to each agent's env file on the computer. Names, ids and statuses only, never a value. `null` when no report line came back from the computer. properties: ok: type: boolean targets: type: array items: $ref: '#/components/schemas/SecretsReportTarget' names: type: object additionalProperties: type: string enum: - cleared - kept - none - would_clear - would_keep description: >- A removal, or its preview: what happened to each name on this computer. `cleared` when at least one copy was taken out, `kept` when a copy stays and none was taken out, `none` when there was no copy. A preview says `would_clear` and `would_keep`. `kept_in` says where a copy stays. kept_in: type: object additionalProperties: type: array items: type: string description: >- A removal, or its preview: where a copy of a name stays. `dotenv` is a line in `/root/.env` from when the computer was created that does not hold your current vault value. An agent id is a line in that agent's file that was changed on the computer, or a file Orgo could not read or write. SecretsReportTarget: type: object description: >- One agent the call reached, or would reach. Agents that are not installed are left out. required: - id - status properties: id: type: string description: >- The agent: `hermes`, `openclaw`, `codex`, `claude-code`, `opencode`, or `agy`. `agents` for the agent step as a whole. status: type: string enum: - updated - unchanged - skipped - shell - cleared - would_update - would_clear description: >- `updated`: the agent's file was written. `unchanged`: on an apply, the file did not change; in a preview, none of the names would go in it (each is in `skipped`). `shell`: the agent has no env file and reads `/root/.env`. `cleared`: Orgo's lines for `names` were taken out. `skipped`: the file was left alone (`reason`). A preview (`dry_run: true`) says `would_update` and `would_clear` in place of `updated` and `cleared`. reason: type: string enum: - owner_managed - unreadable - write_failed - state_unreadable description: >- Why a target was skipped. `owner_managed`: a Hermes whose gateway its owner runs on their own setup. `unreadable`: Orgo can't read the file. `write_failed`: Orgo couldn't write it (never in a preview). `state_unreadable`, on `agents`: `/root/.orgo-secrets.agents.json` does not parse, so no agent file was touched. names: type: array items: type: string description: The names written, taken out, or that would be. skipped: type: array description: Names left out of this file. Never a value. items: type: object required: - name - reason properties: name: type: string reason: type: string enum: - value_format - reserved description: >- `value_format`: the value has a single quote, a backslash, `${`, or a line break of any kind, so it stays out of the agent's file and still reaches `/root/.env`. `reserved` (preview only): the agent keeps its own value for this name, such as Hermes' `HERMES_HOME`, `API_SERVER_` and chat app names, Codex's `CODEX_` names, or OpenClaw's `OPENCLAW_` names. restart: type: string enum: - openclaw-gateway - hermes-gateway description: >- The supervisord program to restart for the change to take effect. `openclaw-gateway` whenever OpenClaw's file changes. `hermes-gateway` when a line leaves Hermes' file: a removal, or an apply that takes out the old line of a value Hermes' file can't hold. responses: Unauthorized: description: 'No usable credential. Send `Authorization: Bearer $ORGO_API_KEY`.' content: application/json: schema: $ref: '#/components/schemas/Error' examples: invalid-key: summary: The key is not one of yours value: error: Invalid API key no-credential: summary: No key and no session value: error: Authentication required AuthUnavailable: description: >- Orgo could not verify the credential because of a server-side fault, such as a database outage. Retry after `Retry-After`. Do not rotate the key. content: application/json: schema: $ref: '#/components/schemas/Error' example: error: >- Service temporarily unavailable. The database is not accepting requests. Retry shortly. headers: Retry-After: description: Seconds to wait before retrying. Always `5`. schema: type: string example: '5' securitySchemes: bearerAuth: type: http scheme: bearer description: API key authentication. Get your key at orgo.ai/workspaces ```` This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.