Connect your AI development stack

Bring project knowledge and real application behavior into expert elicitation, benchmark engineering, and improvement. Evaluate systems in your own environment, give coding agents access to benchmark criteria and results, and connect training workflows to your chosen providers.
Talk to our teamRead integration guides
Code
GitHub
GitLab
Bitbucket
Evals
Braintrust
LangSmith
Arize
Langfuse
Alignment
Teammately
Your experts
Implementation
Codex
Claude Code
Cursor
Devin
Training
Fireworks
Together
Baseten
Amazon SageMaker AI
Google Vertex AI
Microsoft Foundry
01 / Evals

Connect your evals to expert-defined benchmarks

Bring work from Braintrust, Arize, Langfuse, and LangSmith into Teammately. Use selected inputs to build benchmark Cases, or submit outputs and recorded trajectories for rubric evaluation and candidate comparison.
Platforms and tools
  • Braintrust
  • LangSmith
  • Arize
  • Langfuse
Connection methods
  • Selected trace import
  • File upload
  • API
  • CLI
Trace import
Connect Langfuse or LangSmith, browse interactions, and preview selected inputs before importing them into Case Pool for benchmark work.
Case import guide
Outcome submission
Run your application against a Teammately benchmark in your own environment. Submit selected outputs and recorded trajectories from Braintrust, Arize, or your application through file upload, API, or CLI.
You choose which inputs and outcomes to import. Observed outputs remain distinct from expert-approved answers. Choose optional managed Harness execution when you want Trialground to run the candidate.
External output guide
Evals
Braintrust
LangSmith
Arize
Langfuse
Case Pool / Trialground
Benchmark Cases and Run comparisons
Import selected inputs / submit outcomes
  1. Import or select benchmark Cases
  2. Execute in your environment
  3. Submit outcomes
  4. Review completed evaluation
  5. Compare Runs
02 / Project knowledge

Give agents the context behind your application

Connect repositories, websites, and documents as project reference material. Give Teammately agents access to the source context they need when preparing expert contributions and developing benchmark criteria.
Platforms and tools
  • GitHub
  • GitLab
  • Bitbucket
Connection methods
  • Repository connection
  • Website connection
  • File upload
Repository authorization and project selection are separate. Reference material informs project work; experts approve Policies and Rubrics.
Reference Materials guide
Project knowledge
GitHub
GitLab
Bitbucket
WebsitesDocuments
Reference Materials
Project work
Process and index
  1. Authorize a source
  2. Select the project’s repository/ref, website, or files
  3. Process and index
  4. Inspect available reference material
  5. Use it in project work
03 / Coding agents

Bring benchmark criteria into your coding-agent workflow

Give your coding agent the criteria, Cases, and evaluation results needed to improve your application. Keep implementation in your repository and assess the changed behavior against the selected benchmark.
Platforms and tools
  • Codex
  • Claude Code
  • Cursor
  • Devin
Connection methods
  • API
  • CLI
  • MCP
MCP reads benchmark context and results. CLI/API submissions send the outcomes you choose to evaluate.
Evaluation integration guide
Coding agents
Benchmark criteria and results
Your coding agent / repository
Outcomes submitted for evaluation
Platforms and tools
  • Codex
  • Claude Code
  • Cursor
  • Devin
MCP reads / CLI submissions
  1. Select benchmark and optional baseline
  2. Give the task to your agent
  3. Retrieve criteria/results through MCP
  4. Modify and execute customer code
  5. Submit outcomes
  6. Inspect comparison
04 / Application behavior

Connect application behavior to coverage work

Send selected application interactions and feedback to Teammately. Investigate behavior in context and use the findings to guide coverage decisions.
Connection methods
  • API
  • File import
SDKs
  • Python
  • JavaScript / TypeScript
  • Go
Your application selects the records it sends. Investigation and review connect those records to coverage decisions.
Integration documentation
Application behavior
Application interactionsFeedback
Investigation
Coverage decisions
SDKs
  • Python
  • JavaScript / TypeScript
  • Go
Send selected records
  1. Configure an application source
  2. Submit selected records
  3. Inspect processing results
  4. Investigate behavior
  5. Review proposed coverage changes
05 / Training

Connect training data to your training environment

Use expert criteria and coverage objectives to prepare training material. Connect a supported provider to configure and monitor training, then evaluate the candidate through the appropriate model or Harness path.
Platforms and tools
  • Fireworks
  • Together
  • Amazon Bedrock
  • Azure OpenAI
  • Microsoft Foundry
  • Baseten
  • Amazon SageMaker AI
  • Google Vertex AI
Connection methods
  • Dataset download
  • Provider connection
Dataset download is an independent path. Connected training depends on the provider’s supported models, methods, permissions, and configuration.
Integration documentation
Training
Training dataset
Export or connected training
Candidate evaluation
Platforms and tools
  • Fireworks
  • Together
  • Amazon Bedrock
  • Azure OpenAI
  • Microsoft Foundry
  • Baseten
  • Amazon SageMaker AI
  • Google Vertex AI
Choose your path
  1. Prepare and inspect dataset
  2. Choose export or connected training
  3. Configure supported provider settings
  4. Explicitly launch training
  5. Evaluate candidate

Integration directory

Platforms and tools

Evals

Connect your existing eval work to benchmark Cases, rubric evaluation, and candidate comparisons.
Explore the workflow
  • Braintrust
    Submit selected outputs and recorded trajectories through file upload, API, or CLI.
  • LangSmith
    Connect a Workspace and choose its tracing Projects.
  • Arize
    Submit selected outputs and recorded trajectories through file upload, API, or CLI.
  • Langfuse
    Connect the Langfuse Project you want to browse.

Reference sources

Select the source material your project uses as reference knowledge.
Explore the workflow
  • GitHub.com
    Read-only repository contents and metadata through the Teammately GitHub App.
  • GitLab.com
    Select an accessible repository and ref for the project.
  • Bitbucket Cloud
    Select a repository and ref accessible to the connected account.

Training

Configure provider-specific fine-tuning or custom training. Model and method support follows the selected provider and configuration.
Explore the workflow
  • Fireworks
    Connected fine-tuning.
  • Together
    Connected fine-tuning.
  • Amazon Bedrock
    Connected fine-tuning.
  • Azure OpenAI
    Connected fine-tuning.
  • Microsoft Foundry
    Connected fine-tuning.
  • Baseten
    Configured custom training.
  • Amazon SageMaker AI
    Configured custom training.
  • Google Vertex AI
    Configured custom training.

Bring your own models

Evaluate models from your existing stack against your expert-defined benchmarks in Trialground.
Explore the workflow
Example model choices
  • OpenAI
  • Claude
  • Gemini
  • Grok
  • Muse (Meta)
  • Mistral
Plus your own hosted model or application.
Gateways
  • OpenRouter
Your models / application
Teammately
Trialground
Rubric results and comparisons
Connection methods
  • Compatible inference endpoint
    Evaluate a configured model through its HTTPS endpoint. Inference access is separate from training access.
  • Harness
    Run your own hosted model or application through a Harness and evaluate its outputs in Trialground.

Notifications

Connect team messaging to the supported Teammately notification workflow.
  • Slack
    Receive supported product milestone notifications according to your notification preferences.
  • Microsoft Teams
    Install the Teammately app and request a personal test message.

Connection methods

Reference imports

Bring website and file content into project reference knowledge.
Explore the workflow
  • Websites
    Connect selected website sources for processing and indexing.
  • Uploaded files
    Supply documents as project reference material.

Developer tools

Connect benchmark criteria and evaluation results to your own development workflow.
Explore the workflow
  • API
    Import Cases, submit external outcomes, and retrieve evaluation results.
  • CLI
    Retrieve Cases, submit outcomes or supported Harness source, and inspect Run comparisons.
  • MCP
    Read benchmark context, criteria, evaluation results, and comparisons from compatible agent tools.

Application ingestion

Send selected application records and feedback for investigation and coverage work.
Explore the workflow
  • API and file import
    Submit selected records and inspect processing results.
  • Python SDK
    Capture and submit selected records from Python applications.
  • JavaScript / TypeScript SDK
    Capture and submit selected records from server-side JavaScript and TypeScript applications.
  • Go SDK
    Capture and submit selected records from Go applications and workers.

Connect your existing work to expert-grounded improvement.

Teammately connects existing context, execution, and development tools to expert-grounded improvement. Each workflow stands on its own.
Talk to our teamRead integration guides