Cohere
Put Cohere to work on your own records, inside your rules, with a person approving what it changes.
16 ways to read, 7 ways to write, each write switched on by you.
Quality checks
Built and tested with example responses.
- Ready to connect: doneEvery listed action is built in.
- Built and tested with example responses: doneEach action is built from the provider’s published API and tested against example responses, errors, rate limits and paging.
Connected, approved, proven.
The things teams do most with Cohere, an example flow with a person in the loop, and what Nexra keeps as proof.
Top actions
- Create batchApproval by rule
- Create embed jobApproval by rule
- List batchesRead
Example flow
- Cohere · List batchesDone
- Cohere · Create batchNeeds approval
- Nexra · Record the decision and result in the evidence chainProven
Example workflow. Routine runs go through on their own. Only a change your rules flag, such as one over a limit you set, waits for the right person, who sees the exact change and the rule before it reaches Cohere.
What gets logged
- Whether it ran on its own or who approved it, and under which rule
- A fingerprint of the exact request sent to Cohere
- Cohere's reply: what it created or changed
- One entry in the evidence chain, linked to the one before
Try this in the sandbox
Run the whole flow with test data in your sandbox, then connect your own account when you are ready to go live.
For developers
Technical detail
Cohere chat, JSON output, embeddings, rerank and classify, plus datasets, embed jobs, Batch API jobs, fine-tuned model status, tokenization and document parsing (beta).
Cohere API docs- Source actions
- 16
- Destination actions
- 7
- Data types
- 16
What you can map
Reads are always available. Writes are off until you switch them on for a connection, and governed writes follow your policies.
API key1
- Check API keySource
Confirms the key is valid and returns its organization and owner IDs.
Batches3
- Create batchDestinationNot retried automaticallyApproval before production
Runs a validated batch input dataset (for example batch-chat-v2-input) through a model asynchronously.
- Get batchSource
Status and output dataset of one batch.
- List batchesSource
Batch API jobs with status, record counts and token totals.
Chat1
- Generate text (chat)DestinationNot retried automaticallyApproval before production
Generates a reply with a Command model (Chat API v2). Supply a JSON schema to receive structured JSON.
Classification1
- Classify textsDestinationNot retried automaticallyApproval before production
Classifies up to 96 texts with labelled examples or a fine-tuned classifier (Classify API v1).
Dataset usage1
- Get dataset storage usageSource
Bytes of dataset storage used by the organization (limit 10 GB).
Datasets2
- Get datasetSource
One dataset with validation status and parts.
- List datasetsSourceIncremental
Datasets (embed, batch and fine-tune inputs and results) with validation status.
Detokenization1
- Detokenize tokensSource
Converts a model's token IDs back to text.
Embed jobs3
- Create embed jobDestinationNot retried automaticallyApproval before production
Embeds a validated embed-input dataset asynchronously; results land in an embed-result dataset.
- Get embed jobSource
Status and output dataset of one embed job.
- List embed jobsSource
Asynchronous embed jobs with status and input and output datasets.
Embeddings1
- Create embeddingsDestinationNot retried automaticallyApproval before production
Embeds up to 96 texts (Embed API v2); returns one record per text.
Finetuned models2
- Get fine-tuned modelSource
One fine-tuned model with its training settings and status (deprecated API).
- List fine-tuned modelsSource
Fine-tuned models in the organization with base model, dataset, hyperparameters and status (deprecated API).
Finetuning events1
- List fine-tuning eventsSource
Status history of one fine-tuned model (queued, fine-tuning, deploying, ready), newest first.
Finetuning step metrics1
- List fine-tuning training metricsSource
Training-step metrics (accuracy, cross-entropy and similar) recorded while a model was fine-tuned.
Models2
- Get modelSource
Metadata for one model, including endpoints, context length and features.
- List modelsSource
Models available to the API key, optionally filtered to one endpoint.
Parse1
- Parse document image (beta)DestinationNot retried automaticallyApproval before production
Parses a document image at an HTTPS URL into Markdown (Parse API v2, beta); one record per page.
Rerank1
- Rerank documentsDestinationNot retried automaticallyApproval before production
Orders documents by relevance to a query (Rerank API v2); returns one record per ranked document.
Tokenization1
- Tokenize textSource
Splits text into a model's tokens (byte-pair encoding) to estimate usage.
Access and limits
- Authentication
- Cohere API key
- Plans and access
- Trial and production API keys from the Cohere dashboard. Trial keys are rate limited and not for production use. Model availability depends on the account.
- Test environment
- A free trial API key can be used for testing with reduced rate limits.
- Rate limits
- Per-endpoint request limits that differ between trial and production keys; HTTP 429 when exceeded.
- Pagination
- page_token cursors for models, batches and fine-tuning reads; datasets use limit and offset; embed jobs return one page.
- Provider events
- Cohere does not deliver webhooks for these APIs. Flows run on demand, on a schedule, or when a scheduled check finds new or changed records.
Before you build a flow
- Streaming, tool calling, document grounding and image inputs are not exposed; chat returns the final text.
- Structured JSON output cannot be combined with documents or tools, as Cohere documents.
- Dataset upload (multipart) and dataset downloads are not implemented; embed jobs and batches run on datasets uploaded outside Nexra, and part download URLs are never mapped.
- Embed job creation, batch creation and Parse are billable and never retried automatically. Cancelling embed jobs and batches is not implemented.
- Parse (/v2/parse) is a beta API that accepts images only.
- Audio transcription (/v2/audio/transcriptions) needs a multipart file upload and is not implemented.
- Generation, embedding, rerank and classify are billed; they are write operations that must be enabled per connection and are never retried automatically.
Developer reference
- Create the Cohere credentials described above with the narrowest permissions your flows need.
- In Nexra, open Connect, then Connections, then New connection, choose Cohere, enter its settings and credentials, and tick only the write actions you need. Connection guide
- Test the connection, then use its actions as flow sources and destinations. Flow guide
Machine-readable detail: GET /api/public/connect/connectors/cohere and the Connections API.