Can I use Cohere as a CLI?

Yes. KosmoKrator exposes a Cohere CLI through kosmo integrations:call and the provider shortcut command. Commands can return JSON for scripts, CI, and coding agents.

Can I expose Cohere through MCP?

Yes. The KosmoKrator MCP gateway can expose selected Cohere tools to clients such as Claude Code, Cursor, Codex, and other MCP-compatible tools.

Where are Cohere credentials stored?

Cohere credentials are configured in KosmoKrator's local integration settings and reused by CLI, Lua, and MCP gateway surfaces.

Can I deny write access for Cohere MCP tools?

Yes. Start the gateway with --write=deny for read-only clients, or use ask/allow only for trusted workspaces.

data

Cohere CLI for AI Agents

Use the Cohere CLI from KosmoKrator to call Cohere tools headlessly, return JSON, inspect schemas, and automate workflows from coding agents, scripts, and CI.

Cohere CLI Setup

Cohere can be configured headlessly with `kosmokrator integrations:configure cohere`.

Install, configure, and verify

# Install KosmoKrator first if it is not available on PATH.
curl -fsSL https://raw.githubusercontent.com/OpenCompanyApp/kosmokrator/main/install.sh | bash

# Configure and verify this integration.
kosmokrator integrations:configure cohere --set api_key="$COHERE_API_KEY" --enable --read allow --write ask --json
kosmokrator integrations:doctor cohere --json
kosmokrator integrations:status --json

Credentials

Authentication type: API key api_key. Configure credentials once, then reuse the same stored profile from scripts, coding CLIs, Lua, and MCP.

Key	Env var	Type	Required	Label
`api_key`	`COHERE_API_KEY`	Secret `secret`	yes	API Key
`client_name`	`COHERE_CLIENT_NAME`	Text `text`	no	Client Name
`url`	`COHERE_URL`	URL `url`	no	API Base URL

Command Patterns

The generic command is stable across every integration. The provider shortcut is shorter for humans.

Generic CLI call

kosmo integrations:call cohere.cohere_chat '{"model":"example_model","messages":"example_messages","stream":true,"tools":"example_tools","documents":"example_documents","citation_options":"example_citation_options","response_format":"example_response_format","safety_mode":"example_safety_mode"}' --json

Provider shortcut

kosmo integrations:cohere cohere_chat '{"model":"example_model","messages":"example_messages","stream":true,"tools":"example_tools","documents":"example_documents","citation_options":"example_citation_options","response_format":"example_response_format","safety_mode":"example_safety_mode"}' --json

Discovery

These commands return structured output for coding agents that need to inspect capabilities before choosing a function.

Discovery commands

kosmo integrations:docs cohere --json
kosmo integrations:docs cohere.cohere_chat --json
kosmo integrations:schema cohere.cohere_chat --json
kosmo integrations:search "Cohere" --json
kosmo integrations:list --json

Automation Contexts

The same configured command surface works in these environments. The command does not change unless the host wrapper, credentials, or permissions change.

CI Run integration calls from CI jobs with JSON output, explicit credentials, and predictable exit status. Cron Jobs Schedule repeatable integration workflows from cron while keeping credentials in KosmoKrator config. Shell Scripts Call integration functions from shell scripts with stable JSON input and output. Headless Automation Use KosmoKrator as a non-interactive integration runtime for local automations and wrappers. Coding Agents Let coding agents discover schemas and execute integration functions through CLI commands or MCP.

CLI Functions

Every function below can be called headlessly. Commands are highlighted, copyable, and scroll horizontally when payloads are long.

`cohere.cohere_chat`

Generate a non-streaming response with Cohere v2 Chat. Supports messages, documents, tools, citations, JSON response format, safety mode, sampling controls, and reasoning configuration.

Read read

Parameters: model, messages, stream, tools, documents, citation_options, response_format, safety_mode, max_tokens, stop_sequences, temperature, seed, frequency_penalty, presence_penalty, k, p, logprobs, tool_choice, thinking, priority, strict_tools

Generic call

kosmo integrations:call cohere.cohere_chat '{"model":"example_model","messages":"example_messages","stream":true,"tools":"example_tools","documents":"example_documents","citation_options":"example_citation_options","response_format":"example_response_format","safety_mode":"example_safety_mode"}' --json

Shortcut

kosmo integrations:cohere cohere_chat '{"model":"example_model","messages":"example_messages","stream":true,"tools":"example_tools","documents":"example_documents","citation_options":"example_citation_options","response_format":"example_response_format","safety_mode":"example_safety_mode"}' --json

`cohere.cohere_embed`

Create Cohere v2 embeddings for texts, image data URIs, or mixed inputs. Use input_type to match the downstream task such as search_document, search_query, classification, clustering, or image.

Read read

Parameters: model, input_type, texts, images, inputs, max_tokens, output_dimension, embedding_types, truncate, priority

Generic call

kosmo integrations:call cohere.cohere_embed '{"model":"example_model","input_type":"example_input_type","texts":"example_texts","images":"example_images","inputs":"example_inputs","max_tokens":1,"output_dimension":1,"embedding_types":"example_embedding_types"}' --json

Shortcut

kosmo integrations:cohere cohere_embed '{"model":"example_model","input_type":"example_input_type","texts":"example_texts","images":"example_images","inputs":"example_inputs","max_tokens":1,"output_dimension":1,"embedding_types":"example_embedding_types"}' --json

`cohere.cohere_rerank`

Rerank a list of documents for a search query using Cohere v2 Rerank. Documents may be strings; structured data should be converted to YAML strings before calling.

Read read

Parameters: model, query, documents, top_n, max_tokens_per_doc, priority

Generic call

kosmo integrations:call cohere.cohere_rerank '{"model":"example_model","query":"example_query","documents":"example_documents","top_n":1,"max_tokens_per_doc":1,"priority":1}' --json

Shortcut

kosmo integrations:cohere cohere_rerank '{"model":"example_model","query":"example_query","documents":"example_documents","top_n":1,"max_tokens_per_doc":1,"priority":1}' --json

`cohere.cohere_tokenize`

Tokenize text with the tokenizer used by a Cohere model. Use this before budgeting prompts or debugging token boundaries.

Read read

Parameters: text, model

Generic call

kosmo integrations:call cohere.cohere_tokenize '{"text":"example_text","model":"example_model"}' --json

Shortcut

kosmo integrations:cohere cohere_tokenize '{"text":"example_text","model":"example_model"}' --json

`cohere.cohere_detokenize`

Convert Cohere model token IDs back to text with the tokenizer used by a specific model.

Read read

Parameters: tokens, model

Generic call

kosmo integrations:call cohere.cohere_detokenize '{"tokens":"example_tokens","model":"example_model"}' --json

Shortcut

kosmo integrations:cohere cohere_detokenize '{"tokens":"example_tokens","model":"example_model"}' --json

`cohere.cohere_list_models`

List Cohere models with optional pagination and endpoint/default filters.

Read read

Parameters: page_size, page_token, endpoint, default_only

Generic call

kosmo integrations:call cohere.cohere_list_models '{"page_size":1,"page_token":"example_page_token","endpoint":"example_endpoint","default_only":true}' --json

Shortcut

kosmo integrations:cohere cohere_list_models '{"page_size":1,"page_token":"example_page_token","endpoint":"example_endpoint","default_only":true}' --json

`cohere.cohere_get_model`

Get Cohere model metadata including compatible endpoints, deprecation state, context length, features, and sampling defaults.

Read read

Parameters: model

Generic call

kosmo integrations:call cohere.cohere_get_model '{"model":"example_model"}' --json

Shortcut

kosmo integrations:cohere cohere_get_model '{"model":"example_model"}' --json

`cohere.cohere_create_embed_job`

Start a Cohere embed job for a validated embed-input dataset. Use list/get embed job to track completion and read output_dataset_id.

Write write

Parameters: model, dataset_id, input_type, truncate, name

Generic call

kosmo integrations:call cohere.cohere_create_embed_job '{"model":"example_model","dataset_id":"example_dataset_id","input_type":"example_input_type","truncate":"example_truncate","name":"example_name"}' --json

Shortcut

kosmo integrations:cohere cohere_create_embed_job '{"model":"example_model","dataset_id":"example_dataset_id","input_type":"example_input_type","truncate":"example_truncate","name":"example_name"}' --json

`cohere.cohere_list_embed_jobs`

List Cohere embed jobs for the authenticated user.

Read read

Parameters: none

Generic call

kosmo integrations:call cohere.cohere_list_embed_jobs '{}' --json

Shortcut

kosmo integrations:cohere cohere_list_embed_jobs '{}' --json

`cohere.cohere_get_embed_job`

Get details for a Cohere embed job, including status and output_dataset_id when complete.

Read read

Parameters: job_id

Generic call

kosmo integrations:call cohere.cohere_get_embed_job '{"job_id":"example_job_id"}' --json

Shortcut

kosmo integrations:cohere cohere_get_embed_job '{"job_id":"example_job_id"}' --json

`cohere.cohere_cancel_embed_job`

Cancel an active Cohere embed job. Cohere may bill for work already processed, and partial results are not returned.

Write write

Parameters: job_id

Generic call

kosmo integrations:call cohere.cohere_cancel_embed_job '{"job_id":"example_job_id"}' --json

Shortcut

kosmo integrations:cohere cohere_cancel_embed_job '{"job_id":"example_job_id"}' --json

`cohere.cohere_create_dataset`

Upload a dataset file to Cohere. For embed jobs, use type=embed-input and a JSONL/CSV/TXT file that matches Cohere dataset rules.

Write write

Parameters: name, type, filename, content, eval_filename, eval_content, keep_original_file, skip_malformed_input, keep_fields, optional_fields, text_separator, csv_delimiter

Generic call

kosmo integrations:call cohere.cohere_create_dataset '{"name":"example_name","type":"example_type","filename":"example_filename","content":"example_content","eval_filename":"example_eval_filename","eval_content":"example_eval_content","keep_original_file":true,"skip_malformed_input":true}' --json

Shortcut

kosmo integrations:cohere cohere_create_dataset '{"name":"example_name","type":"example_type","filename":"example_filename","content":"example_content","eval_filename":"example_eval_filename","eval_content":"example_eval_content","keep_original_file":true,"skip_malformed_input":true}' --json

`cohere.cohere_list_datasets`

List Cohere datasets with optional datasetType, before, after, limit, and offset filters.

Read read

Parameters: datasetType, before, after, limit, offset

Generic call

kosmo integrations:call cohere.cohere_list_datasets '{"datasetType":"example_datasetType","before":"example_before","after":"example_after","limit":1,"offset":1}' --json

Shortcut

kosmo integrations:cohere cohere_list_datasets '{"datasetType":"example_datasetType","before":"example_before","after":"example_after","limit":1,"offset":1}' --json

`cohere.cohere_get_dataset`

Get a Cohere dataset by ID, including validation status, schema, dataset parts, and metrics.

Read read

Parameters: dataset_id

Generic call

kosmo integrations:call cohere.cohere_get_dataset '{"dataset_id":"example_dataset_id"}' --json

Shortcut

kosmo integrations:cohere cohere_get_dataset '{"dataset_id":"example_dataset_id"}' --json

`cohere.cohere_delete_dataset`

Delete a Cohere dataset by ID. Cohere also automatically expires datasets after its retention period.

Write write

Parameters: dataset_id

Generic call

kosmo integrations:call cohere.cohere_delete_dataset '{"dataset_id":"example_dataset_id"}' --json

Shortcut

kosmo integrations:cohere cohere_delete_dataset '{"dataset_id":"example_dataset_id"}' --json

`cohere.cohere_get_dataset_usage`

Get Cohere organization dataset storage usage in bytes.

Read read

Parameters: none

Generic call

kosmo integrations:call cohere.cohere_get_dataset_usage '{}' --json

Shortcut

kosmo integrations:cohere cohere_get_dataset_usage '{}' --json

`cohere.cohere_create_audio_transcription`

Transcribe an audio file with Cohere v2 Audio Transcriptions. Provide file content directly; supported extensions include flac, mp3, mpeg, mpga, ogg, and wav.

Read read

Parameters: filename, content, model, language, temperature

Generic call

kosmo integrations:call cohere.cohere_create_audio_transcription '{"filename":"example_filename","content":"example_content","model":"example_model","language":"example_language","temperature":1}' --json

Shortcut

kosmo integrations:cohere cohere_create_audio_transcription '{"filename":"example_filename","content":"example_content","model":"example_model","language":"example_language","temperature":1}' --json

`cohere.cohere_classify`

Classify text with Cohere v1 Classify. Upstream marks this endpoint deprecated; prefer newer chat or embedding workflows unless you need legacy classify compatibility.

Read read

Parameters: inputs, examples, model, truncate, preset

Generic call

kosmo integrations:call cohere.cohere_classify '{"inputs":"example_inputs","examples":"example_examples","model":"example_model","truncate":"example_truncate","preset":"example_preset"}' --json

Shortcut

kosmo integrations:cohere cohere_classify '{"inputs":"example_inputs","examples":"example_examples","model":"example_model","truncate":"example_truncate","preset":"example_preset"}' --json

Function Schemas

Use these parameter tables when building CLI payloads without calling integrations:schema first.

cohere.cohere_chat 21 parameters

Schema command

kosmo integrations:schema cohere.cohere_chat --json

Parameter	Type	Required	Description
`model`	string	yes	Cohere chat model ID, for example command-a-03-2025.
`messages`	array	yes	Chronological chat messages with user, assistant, system, or tool roles.
`stream`	boolean	no	Must be false or omitted. Streaming SSE is not supported by this tool.
`tools`	array	no	Tool/function definitions available to the model.
`documents`	array	no	Relevant documents as strings or objects for citation-aware generation.
`citation_options`	object	no	Citation generation options.
`response_format`	object	no	Response format control, such as json_object with optional JSON schema.
`safety_mode`	string	no	Safety instruction mode for compatible models.
`max_tokens`	integer	no	Maximum output tokens.
`stop_sequences`	array	no	Up to five stop strings.
`temperature`	number	no	Sampling temperature.
`seed`	integer	no	Best-effort deterministic seed.
`frequency_penalty`	number	no	Penalty for repeated token frequency.
`presence_penalty`	number	no	Penalty for already-present tokens.
`k`	integer	no	Top-k sampling control.
`p`	number	no	Top-p sampling control.
`logprobs`	boolean	no	Include token log probabilities when supported.
`tool_choice`	string	no	Force tool use or direct response for compatible models.
`thinking`	object	no	Reasoning feature configuration for supported models.
`priority`	integer	no	Priority from 0 to 999, where lower values are handled earlier.
`strict_tools`	boolean	no	Force tool calls to follow the supplied tool schema strictly.

cohere.cohere_embed 10 parameters

Schema command

kosmo integrations:schema cohere.cohere_embed --json

Parameter	Type	Required	Description
`model`	string	yes	Embedding model ID.
`input_type`	string	yes	Embedding input type.
`texts`	array	no	Text inputs to embed. Maximum 96 per call.
`images`	array	no	Image data URI inputs. Maximum 1 image per call.
`inputs`	array	no	Mixed text/image component inputs. Maximum 96 per call.
`max_tokens`	integer	no	Maximum tokens to embed per input.
`output_dimension`	integer	no	Output embedding dimension for embed-v4 and newer models.
`embedding_types`	array	no	Embedding formats such as float, int8, uint8, binary, or ubinary.
`truncate`	string	no	How to truncate inputs that exceed context.
`priority`	integer	no	Priority from 0 to 999, where lower values are handled earlier.

cohere.cohere_rerank 6 parameters

Schema command

kosmo integrations:schema cohere.cohere_rerank --json

Parameter	Type	Required	Description
`model`	string	yes	Rerank model ID, for example rerank-v4.0-pro.
`query`	string	yes	Search query.
`documents`	array	yes	Documents to compare to the query. Cohere recommends no more than 1000 per call.
`top_n`	integer	no	Maximum number of ranked results to return.
`max_tokens_per_doc`	integer	no	Token truncation limit per document. Defaults to 4096 upstream.
`priority`	integer	no	Priority from 0 to 999, where lower values are handled earlier.

cohere.cohere_tokenize 2 parameters

Schema command

kosmo integrations:schema cohere.cohere_tokenize --json

Parameter	Type	Required	Description
`text`	string	yes	Text to tokenize.
`model`	string	yes	Model whose tokenizer should be used.

cohere.cohere_detokenize 2 parameters

Schema command

kosmo integrations:schema cohere.cohere_detokenize --json

Parameter	Type	Required	Description
`tokens`	array	yes	Token integer IDs to detokenize.
`model`	string	yes	Model whose tokenizer should be used.

cohere.cohere_list_models 4 parameters

Schema command

kosmo integrations:schema cohere.cohere_list_models --json

Parameter	Type	Required	Description
`page_size`	number	no	Maximum models per page. Range: 1-1000. Default: 20.
`page_token`	string	no	Token from next_page_token in a previous response.
`endpoint`	string	no	Only return models compatible with this endpoint.
`default_only`	boolean	no	Only return default models for the endpoint. Valid only when endpoint is provided.

cohere.cohere_get_model 1 parameters

Schema command

kosmo integrations:schema cohere.cohere_get_model --json

Parameter	Type	Required	Description
`model`	string	yes	Model name to retrieve.

cohere.cohere_create_embed_job 5 parameters

Schema command

kosmo integrations:schema cohere.cohere_create_embed_job --json

Parameter	Type	Required	Description
`model`	string	yes	Embedding model ID.
`dataset_id`	string	yes	Validated embed-input dataset ID.
`input_type`	string	yes	Embedding input type.
`truncate`	string	no	How to truncate text that exceeds model limits.
`name`	string	no	Optional embed job display name.

cohere.cohere_list_embed_jobs 0 parameters

Schema command

kosmo integrations:schema cohere.cohere_list_embed_jobs --json

Parameter	Type	Required	Description
No parameters.

cohere.cohere_get_embed_job 1 parameters

Schema command

kosmo integrations:schema cohere.cohere_get_embed_job --json

Parameter	Type	Required	Description
`job_id`	string	yes	Embed job ID.

cohere.cohere_cancel_embed_job 1 parameters

Schema command

kosmo integrations:schema cohere.cohere_cancel_embed_job --json

Parameter	Type	Required	Description
`job_id`	string	yes	Embed job ID to cancel.

cohere.cohere_create_dataset 12 parameters

Schema command

kosmo integrations:schema cohere.cohere_create_dataset --json

Parameter	Type	Required	Description
`name`	string	yes	Dataset display name.
`type`	string	yes	Dataset type, for example embed-input.
`filename`	string	yes	Filename for the uploaded dataset.
`content`	string	yes	Raw dataset file content.
`eval_filename`	string	no	Optional evaluation filename.
`eval_content`	string	no	Optional evaluation file content.
`keep_original_file`	boolean	no	Store the original uploaded file.
`skip_malformed_input`	boolean	no	Drop malformed rows instead of failing validation.
`keep_fields`	array	no	Required fields to preserve.
`optional_fields`	array	no	Optional fields to preserve.
`text_separator`	string	no	Separator for raw text uploads.
`csv_delimiter`	string	no	Delimiter for CSV uploads.

cohere.cohere_list_datasets 5 parameters

Schema command

kosmo integrations:schema cohere.cohere_list_datasets --json

Parameter	Type	Required	Description
`datasetType`	string	no	Optional dataset type filter. Use Cohere casing: datasetType.
`before`	string	no	Return datasets before this ISO date-time.
`after`	string	no	Return datasets after this ISO date-time.
`limit`	number	no	Maximum number of results.
`offset`	number	no	Offset into the result set.

cohere.cohere_get_dataset 1 parameters

Schema command

kosmo integrations:schema cohere.cohere_get_dataset --json

Parameter	Type	Required	Description
`dataset_id`	string	yes	Dataset ID.

cohere.cohere_delete_dataset 1 parameters

Schema command

kosmo integrations:schema cohere.cohere_delete_dataset --json

Parameter	Type	Required	Description
`dataset_id`	string	yes	Dataset ID to delete.

cohere.cohere_get_dataset_usage 0 parameters

Schema command

kosmo integrations:schema cohere.cohere_get_dataset_usage --json

Parameter	Type	Required	Description
No parameters.

cohere.cohere_create_audio_transcription 5 parameters

Schema command

kosmo integrations:schema cohere.cohere_create_audio_transcription --json

Parameter	Type	Required	Description
`filename`	string	yes	Audio filename including extension.
`content`	string	yes	Raw audio file bytes as a string from the host.
`model`	string	yes	Transcription model ID, for example cohere-transcribe-03-2026.
`language`	string	yes	ISO-639-1 language code, for example en.
`temperature`	number	no	Optional sampling temperature between 0 and 1.

cohere.cohere_classify 5 parameters

Schema command

kosmo integrations:schema cohere.cohere_classify --json

Parameter	Type	Required	Description
`inputs`	array	yes	Texts to classify. Maximum 96.
`examples`	array	no	Optional examples as objects with text and label. Not required for fine-tuned classify models.
`model`	string	no	Optional fine-tuned Classify model ID.
`truncate`	string	no	How to handle inputs longer than model limits.
`preset`	string	no	Deprecated upstream preset parameter.

Permissions

Headless calls still follow the integration read/write permission policy. Configure read/write defaults with integrations:configure. Add --force only for trusted automation that should bypass that policy.