Software we use, published for you to use too

A multi-provider AI research workbench with a web interface. It connects to models from Anthropic, AWS Bedrock, Google Gemini, Ollama and X.AI, and adds MCP tool integration, persistent semantic memory, context management and scheduled autonomous actions.

• MCP tool servers over stdio, HTTP or SSE, with per-conversation, per-tool approval
• Persistent memory with semantic search and automatic recall
• Scheduled and one-off autonomous actions, with run history and token usage
• API keys held in the operating system keychain, never in configuration files
• Prompt inspection for threat patterns, and a lockable settings page
• Signed installers for macOS and Windows, or install from PyPI

pip install cognisn-spark

A Python framework for Model Context Protocol servers. Large tool results are spooled into an ephemeral SQLite store and explored by the model through query, aggregate and distinct tools, rather than flooding its context window.

• Response spooling: summaries and a handle instead of 100,000 token dumps
• Hierarchical tool prefixing, so tools from several servers never collide
• Skills guides served on demand, keeping tool descriptions lean
• Strict argument validation with structured, self-correcting errors
• stdio, streamable HTTP and SSE transports, with bearer token authentication

pip install cognisn-spindl

The foundations most Python applications need, done once: layered settings, pluggable secrets backends and run-scoped logging, with an optional lifecycle context manager that ties them together. Spark uses it for keychain-backed secret storage.

• Layered settings with system, user, environment and runtime precedence
• Secrets in the OS keyring, an AES-encrypted file, or AWS Secrets Manager
• Run-scoped logging with retention, structured JSON output and stdio-safe mode
• Platform-aware configuration, data and log locations on macOS, Linux and Windows

pip install cognisn-konfig

Deploy complete Jira Cloud projects, including issue types, workflows and screens, from YAML templates through the REST API v3. Every resource created is tracked, so a deployment can be rolled back precisely.

• Whole-project deployment from version-controlled YAML templates
• Deployment tracking of every resource created
• Precise rollback for test projects or failed deployments
• A simple Python wrapper around the Jira Cloud REST API v3

pip install git+https://github.com/Cognisn/jirakit.git